AI changes a legal practice differently from a sales team, a foundation or a utility. Start with the work people do, the decisions they carry and what a mistake would cost.
See how a proven workflow scales to your whole team
An assistant can use AI to prepare replies, document checks and scheduling. Legal advice and approval stay with the responsible professional, and checking time counts against the value.
A client needs an answer, a document needs checking and a meeting needs arranging. A small legal practice depends on all of that work reaching the right person.
AI affects these activities differently. Start with the work and its responsibility before deciding where an assistant belongs.
A legal administrative assistant prepares client updates, summarizes matter documents and keeps appointments moving. They also listen when a client is worried and bring urgent concerns to the right professional.
This example follows administrative work in a small practice. Lawyers and licensed paralegals retain the legal advice and decisions; the assistant helps the work reach them ready for review.
Illustrative legal administrative work informed by O*NET and Law Society of Ontario guidance. It is not a client result.
In this example practice, shared documents and calendars replaced retyping dictated letters, carrying paper files, recopying appointments and collating duplicate bundles. Other practices may still need some of that work.
Managing portal uploads and tracking electronic signatures became part of the day. Those responsibilities came with digital systems, before adding AI to the process.
Listening to a worried client and clarifying incomplete instructions depend on understanding what the person means. The assistant knows when to pause and involve the professional responsible for the matter.
The assistant escalates urgent concerns to the responsible professional and coordinates planned handovers with an agreed recipient. Resolving diary clashes also needs someone who can weigh conflicting commitments.
AI can draft a client status update, summarize documents or prepare a routine request for missing material. The assistant checks facts and routes legal content to the responsible professional before use.
AI can compare received documents with an approved checklist and prepare portal uploads or signature follow-ups. The assistant confirms missing items, the matter and the recipient before acting.
Approved correspondence can be filed against a confirmed matter, and reminders can follow an agreed appointment. Rules may already handle this; AI is useful where interpreting the incoming material adds something.
Receipt matching needs a confirmed matter identifier and a check on the result. Missing or conflicting information goes back to a person; the workflow does not decide legal deadlines, move money or approve a filing.
Checking an AI draft means verifying its claims against actual sources and catching invented or misattributed material. This differs from checking whether the matter contains the documents on an approved checklist.
Maintaining approved examples is another job. When a check exposes a recurring error, the practice updates its instructions and examples, then checks whether later work improves.
Less preparation could leave more time for client updates, missing information and difficult handovers. The useful result is a matter that moves forward with fewer avoidable delays.
That benefit has to survive checking and corrections. Measure the work accepted by the practice, including the new review and maintenance tasks, before treating the change as extra capacity.
Visual size is illustrative, not a measured productivity multiplier. Any extra capacity still needs evidence.
Routine reminders may be ready for wider use while unfamiliar documents still need close preparation and review. The same level of AI use will not fit every matter or every task.
The proportions here illustrate a possible allocation, not measured adoption or time saved. Start with approved material and a bounded workflow, then expand where quality, permissions and the team support it.
Compare the preparation avoided with time spent checking drafts, correcting errors and keeping examples current. Include software use, setup and training before deciding whether the change pays for itself.
This is a planning example, not a savings claim for a real practice. A useful pilot establishes the baseline and the accepted result so the next decision rests on evidence.
Net capacity: Measure useful time. Subtract checking, corrections and maintenance from preparation time avoided.
Per accepted result: Include review. A quick draft only helps when the practice can accept it.
Operating cost: Track each run. Include software use and the work needed to keep it useful.
Payback: Test repeat use. Compare repeated benefit with setup and training.
Original MTW scenario informed by occupational duties; no source supplies AI allocations or savings.
A client needs an answer, a document needs checking and a meeting needs arranging. A small legal practice depends on all of that work reaching the right person.
AI affects these activities differently. Start with the work and its responsibility before deciding where an assistant belongs.
Illustration: a broad field represents work across a small legal practice. It does not represent staff numbers, hours or savings.
A legal administrative assistant prepares client updates, summarizes matter documents and keeps appointments moving. They also listen when a client is worried and bring urgent concerns to the right professional.
This example follows administrative work in a small practice. Lawyers and licensed paralegals retain the legal advice and decisions; the assistant helps the work reach them ready for review.
Illustrative legal administrative work informed by O*NET and Law Society of Ontario guidance. It is not a client result.
Illustration: sixteen historical tasks form the administrative assistant’s starting bundle, including client support, document preparation, records and paper routines. Legal advice and decisions remain with the responsible professional.
In this example practice, shared documents and calendars replaced retyping dictated letters, carrying paper files, recopying appointments and collating duplicate bundles. Other practices may still need some of that work.
Managing portal uploads and tracking electronic signatures became part of the day. Those responsibilities came with digital systems, before adding AI to the process.
Illustration: four paper routines leave this example and two digital responsibilities enter. Fourteen tasks remain; this is the example practice’s history, not a claim about every practice.
Listening to a worried client and clarifying incomplete instructions depend on understanding what the person means. The assistant knows when to pause and involve the professional responsible for the matter.
The assistant escalates urgent concerns to the responsible professional and coordinates planned handovers with an agreed recipient. Resolving diary clashes also needs someone who can weigh conflicting commitments.
Illustration: listening to clients, clarifying instructions, resolving diary clashes, escalating concerns and coordinating handovers remain human-led in this scenario. The classification assigns responsibility, not a limit on all possible assistance.
AI can draft a client status update, summarize documents or prepare a routine request for missing material. The assistant checks facts and routes legal content to the responsible professional before use.
AI can compare received documents with an approved checklist and prepare portal uploads or signature follow-ups. The assistant confirms missing items, the matter and the recipient before acting.
Illustration: six tasks receive assistance: client updates, document summaries, routine correspondence, completeness checks, portal uploads and signature tracking. A person checks and accepts the work.
Approved correspondence can be filed against a confirmed matter, and reminders can follow an agreed appointment. Rules may already handle this; AI is useful where interpreting the incoming material adds something.
Receipt matching needs a confirmed matter identifier and a check on the result. Missing or conflicting information goes back to a person; the workflow does not decide legal deadlines, move money or approve a filing.
Illustration: three bounded routines are candidates for automation: approved filing, confirmed reminders and receipt matching. Missing or conflicting information returns to a person.
Checking an AI draft means verifying its claims against actual sources and catching invented or misattributed material. This differs from checking whether the matter contains the documents on an approved checklist.
Maintaining approved examples is another job. When a check exposes a recurring error, the practice updates its instructions and examples, then checks whether later work improves.
Illustration: checking AI drafts against sources and maintaining approved examples add two responsibilities. The final balance is five human-led, six assisted, three automated and two new tasks.
Less preparation could leave more time for client updates, missing information and difficult handovers. The useful result is a matter that moves forward with fewer avoidable delays.
That benefit has to survive checking and corrections. Measure the work accepted by the practice, including the new review and maintenance tasks, before treating the change as extra capacity.
Visual size is illustrative, not a measured productivity multiplier. Any extra capacity still needs evidence.
Illustration: seven originally assisted or automated tasks receive visual enlargement. Their size is emphasis, not a productivity multiplier or a measure of time saved.
Routine reminders may be ready for wider use while unfamiliar documents still need close preparation and review. The same level of AI use will not fit every matter or every task.
The proportions here illustrate a possible allocation, not measured adoption or time saved. Start with approved material and a bounded workflow, then expand where quality, permissions and the team support it.
Illustration: assumed use divides seven eligible tasks: client updates 55%, document summaries 45%, correspondence 60%, completeness checks 50%, approved filing 75%, reminders 80% and receipt matching 65%. These are shares of task instances using assistance, not measured adoption or savings.
Compare the preparation avoided with time spent checking drafts, correcting errors and keeping examples current. Include software use, setup and training before deciding whether the change pays for itself.
This is a planning example, not a savings claim for a real practice. A useful pilot establishes the baseline and the accepted result so the next decision rests on evidence.
Illustration: the tasks and decorative cells form a dollar sign behind four qualitative planning panels. The symbol provides no price, staffing reduction, savings estimate or payback period.
Measure useful time
Net capacity
Subtract checking, corrections and maintenance from preparation time avoided.
Include review
Per accepted result
A quick draft only helps when the practice can accept it.
Track each run
Operating cost
Include software use and the work needed to keep it useful.
Test repeat use
Payback
Compare repeated benefit with setup and training.
Original MTW scenario informed by occupational duties; no source supplies AI allocations or savings.
AI can assemble material for calls and proposals. Understanding the buyer and deciding what the business can promise stay with a person, and checking claims is part of the work.
A buyer needs a clear answer, a proposal needs work and an existing account needs attention. A sales team depends on those activities joining up, not just on sending more messages.
AI may help prepare information or handle a bounded routine. The useful question is where that helps a representative make a better commitment to the customer.
The account representative prepares for calls, answers buyer questions and works through a proposal. They also negotiate what the business can commit to and stay involved when delivery raises a concern.
This example follows a business services account representative. Finance approves credit and exceptional discounts; the delivery owner confirms capacity and the final handoff.
Original MTW scenario informed by O*NET sales duties. AI allocation, earlier methods and use shares are illustrative, not a client result.
Shared customer records replaced paper contact cards and recopying call logs in this example team. Online proposals and service pages replaced its standard fax quotes and mailed brochures.
Keeping CRM opportunities current and running remote demonstrations became ordinary work. The representative still has to judge the next stage and respond to the buyer during a demonstration.
Clarifying buyer priorities and judging service fit take a conversation about what matters. The representative can decide that the service is wrong for the buyer, even when the prospect looks promising.
Negotiating terms, resolving account concerns and agreeing the delivery handoff remain human responsibilities. The representative works within delegated authority and involves the person who can approve an exception.
AI can prepare an account brief and a proposal from permitted customer records and approved service material. The representative checks the sources and decides what belongs in the conversation.
AI can also draft an answer to a buyer question or compare renewal options. The representative checks current terms and fit; an authorized person approves any exception before the message reaches the buyer.
An approved meeting note can become a structured account record, and an enquiry with a confirmed customer ID can reach its existing owner. Missing identifiers or conflicting details go to the sales team for review.
An account activity digest can collect confirmed events with links back to them. Existing rules may do this work; AI adds value only where extracting detail helps. Checks keep it from changing deal stages or sending offers.
Checking AI claims becomes a distinct responsibility. A persuasive proposal is still unusable if it invents a capability, repeats an expired offer or borrows another customer’s information.
The team also maintains tested sales examples. When a proposal repeats a bad claim, the representative supplies the correction and a sales owner approves it before the examples are reused.
Less account preparation could leave more time to clarify buyer priorities and resolve a difficult concern. The test is whether customers get a useful answer without promises the delivery team cannot keep.
Checking claims and maintaining tested examples use time too. Compare preparation avoided with the whole review burden before treating the change as capacity for more accounts.
A familiar service question may suit assistance more often than an unusual renewal. The team should expand use where approved material, account permissions and an accountable reviewer are available.
The proportions here are illustrative shares of task instances using assistance. They are not measured adoption or time saved, and they do not predict a higher win rate.
Count the effort to prepare, review and correct an accepted proposal or answer. Include the new claim checks, tested examples, software use and setup before judging whether the change earns its place.
This example claims no revenue uplift or headcount saving. A pilot needs a baseline and repeatable results before the business can decide how much more work to put through it.
Net account capacity: Count useful time. Subtract claim checks and corrections from account preparation avoided.
Per accepted answer: Include the review. An answer is useful only when its facts and commercial terms are accepted.
Operating cost: Track repeated work. Include tool use and the effort to keep the sales material current.
Payback: Test another account. Compare repeat benefit with setup and training before wider use.
Original MTW scenario informed by O*NET occupational duties; the sources do not supply AI allocations, savings or sales results.
A buyer needs a clear answer, a proposal needs work and an existing account needs attention. A sales team depends on those activities joining up, not just on sending more messages.
AI may help prepare information or handle a bounded routine. The useful question is where that helps a representative make a better commitment to the customer.
Illustration: a wide field represents different activities across a sales team. It does not represent staff numbers, hours or revenue.
The account representative prepares for calls, answers buyer questions and works through a proposal. They also negotiate what the business can commit to and stay involved when delivery raises a concern.
This example follows a business services account representative. Finance approves credit and exceptional discounts; the delivery owner confirms capacity and the final handoff.
Original MTW scenario informed by O*NET sales duties. AI allocation, earlier methods and use shares are illustrative, not a client result.
Illustration: sixteen historical tasks form the representative’s starting bundle. This includes relationship decisions, preparation, record work and earlier paper routines.
Shared customer records replaced paper contact cards and recopying call logs in this example team. Online proposals and service pages replaced its standard fax quotes and mailed brochures.
Keeping CRM opportunities current and running remote demonstrations became ordinary work. The representative still has to judge the next stage and respond to the buyer during a demonstration.
Illustration: four earlier routines leave during the change, and two digital responsibilities enter. Fourteen tasks remain; this is the example team’s history, not a claim that paper selling has disappeared.
Clarifying buyer priorities and judging service fit take a conversation about what matters. The representative can decide that the service is wrong for the buyer, even when the prospect looks promising.
Negotiating terms, resolving account concerns and agreeing the delivery handoff remain human responsibilities. The representative works within delegated authority and involves the person who can approve an exception.
Illustration: five human-led responsibilities receive emphasis. The CRM and remote demonstration tasks also remain unchanged; the four legend categories classify work rather than people.
AI can prepare an account brief and a proposal from permitted customer records and approved service material. The representative checks the sources and decides what belongs in the conversation.
AI can also draft an answer to a buyer question or compare renewal options. The representative checks current terms and fit; an authorized person approves any exception before the message reaches the buyer.
Illustration: four preparation tasks enter the augmented category. The earlier digital responsibilities keep their unchanged category.
An approved meeting note can become a structured account record, and an enquiry with a confirmed customer ID can reach its existing owner. Missing identifiers or conflicting details go to the sales team for review.
An account activity digest can collect confirmed events with links back to them. Existing rules may do this work; AI adds value only where extracting detail helps. Checks keep it from changing deal stages or sending offers.
Illustration: three record-handling routines enter the automated category. These are bounded proposed workflows with source checks and a human exception route.
Checking AI claims becomes a distinct responsibility. A persuasive proposal is still unusable if it invents a capability, repeats an expired offer or borrows another customer’s information.
The team also maintains tested sales examples. When a proposal repeats a bad claim, the representative supplies the correction and a sales owner approves it before the examples are reused.
Illustration: two AI-specific responsibilities enter the new category. The final task balance is seven unchanged, four augmented, three automated and two new.
Less account preparation could leave more time to clarify buyer priorities and resolve a difficult concern. The test is whether customers get a useful answer without promises the delivery team cannot keep.
Checking claims and maintaining tested examples use time too. Compare preparation avoided with the whole review burden before treating the change as capacity for more accounts.
Illustration: seven augmented or automated task tiles receive visual enlargement. Their size is emphasis, not a productivity multiplier or a statement about time spent.
A familiar service question may suit assistance more often than an unusual renewal. The team should expand use where approved material, account permissions and an accountable reviewer are available.
The proportions here are illustrative shares of task instances using assistance. They are not measured adoption or time saved, and they do not predict a higher win rate.
Illustration: the seven eligible tasks split by explicitly assumed use of assistance. Call briefs: 35%; proposals: 45%; buyer answers: 55%; renewal comparisons: 30%; meeting facts: 65%; enquiry routing: 60%; activity digests: 70%. These are scenario inputs, not measured savings.
Count the effort to prepare, review and correct an accepted proposal or answer. Include the new claim checks, tested examples, software use and setup before judging whether the change earns its place.
This example claims no revenue uplift or headcount saving. A pilot needs a baseline and repeatable results before the business can decide how much more work to put through it.
Illustration: the remaining tasks and extra decorative cells form a dollar sign behind four qualitative planning panels. The symbol supplies no revenue, price or payback estimate.
Count useful time
Net account capacity
Subtract claim checks and corrections from account preparation avoided.
Include the review
Per accepted answer
An answer is useful only when its facts and commercial terms are accepted.
Track repeated work
Operating cost
Include tool use and the effort to keep the sales material current.
Test another account
Payback
Compare repeat benefit with setup and training before wider use.
Original MTW scenario informed by O*NET occupational duties; the sources do not supply AI allocations, savings or sales results.
AI can organize applications and draft routine replies. The coordinator hears what a form misses; programme leaders decide awards and finance authorizes payments.
An applicant needs help, a review needs preparation and a funded project needs attention. A foundation relies on those activities joining up throughout a grant.
AI may help with preparation or a bounded routine. Its value depends on whether staff can use the result while keeping responsibility for the funding decision.
A grants program coordinator helps applicants through the process and prepares briefs for review. They also discuss progress with grantees and prepare the next meeting.
This example follows the grantmaking side of a nonprofit foundation. Programme leaders and committees decide awards and material changes; finance authorizes payments. The coordinator helps that work stay connected.
Original MTW planning example informed by occupational duties. Earlier methods and AI use are illustrative; no client result or savings is claimed.
In this example, online forms replaced posted application packs and re-keying. Permissioned review spaces and structured reports replaced paper binders and hand tallies. These changes preceded AI.
Maintaining the grant portal and reconciling dashboards became ordinary work. Staff still test forms against programme rules and investigate discrepancies in the records.
Helping an applicant and clarifying missing facts requires listening. A sparse application may need a question, not an assumption about the organization behind it.
The coordinator discusses grant progress, works through change requests and agrees review priorities with the programme team. A programme leader approves a material change; a tidy summary cannot make that decision.
AI can prepare an application brief and draft an applicant reply from permitted records and approved guidance. The coordinator checks the facts and sends the reply; the reviewer judges the application.
AI can also summarize a grantee report or prepare meeting notes. The coordinator checks reported progress against the source, and attendees confirm decisions before the notes become the official record.
A routine enquiry can reach an internal queue, and an approved document can be indexed against a confirmed grant. AI can interpret wording; ordinary rules enforce identity and access. Ambiguous records return to the coordinator.
A bounded extraction can place stated report figures in a staging table with page references. Checks test units, periods and source matches; uncertain values return to staff. This does not approve a payment or establish the grant’s impact.
Checking AI summaries becomes a separate responsibility. The coordinator compares the draft with the original submission and records unsupported claims or missing context.
Someone must also maintain the AI checks and examples. When guidance changes or an error repeats, the programme owner accepts the correction before the workflow is reused.
Less time preparing briefs and report summaries could leave more attention for applicant questions and progress conversations. That is a possible use of capacity, not a promise that the foundation can handle more grants.
Source checking, corrections and maintaining AI checks take time too. Measure the preparation avoided against that extra work before treating a faster draft as a gain.
A familiar enquiry may suit a routine more often than a nuanced progress report. Thin evidence, conflicting records or an unusual change request needs staff attention.
The proportions here are illustrative shares of task instances using assistance. They are not measured adoption or time saved. Test the records and checks before extending the workflow to another programme.
Compare the effort to prepare an accepted brief or report summary with the full effort to check and correct it. A draft that staff cannot use does not count as a useful result.
Include software use, setup, training and maintained AI checks. This planning example claims no savings; repeated accepted work must justify the ongoing effort before the foundation expands it.
Net coordination capacity: Measure useful time. Planning measure: subtract checking and upkeep from preparation avoided.
Per accepted result: Count the checks. Planning measure: include rejected attempts and the work to correct them.
Operating cost: Track the workflow. Planning measure: include use, integration and ongoing support.
Payback: Test repeat use. Planning measure: compare recurring benefit with setup and training.
Original MTW scenario informed by primary foundation role descriptions and grant administration sources; no source supplies AI allocations or savings.
An applicant needs help, a review needs preparation and a funded project needs attention. A foundation relies on those activities joining up throughout a grant.
AI may help with preparation or a bounded routine. Its value depends on whether staff can use the result while keeping responsibility for the funding decision.
Illustration: the broad field represents activities across a foundation, not staff numbers, time or grant value.
A grants program coordinator helps applicants through the process and prepares briefs for review. They also discuss progress with grantees and prepare the next meeting.
This example follows the grantmaking side of a nonprofit foundation. Programme leaders and committees decide awards and material changes; finance authorizes payments. The coordinator helps that work stay connected.
Original MTW planning example informed by occupational duties. Earlier methods and AI use are illustrative; no client result or savings is claimed.
Illustration: sixteen tasks form a historical starting bundle, including applicant support, preparation, records and paper routines.
In this example, online forms replaced posted application packs and re-keying. Permissioned review spaces and structured reports replaced paper binders and hand tallies. These changes preceded AI.
Maintaining the grant portal and reconciling dashboards became ordinary work. Staff still test forms against programme rules and investigate discrepancies in the records.
Illustration: four historical routines leave during this step and two digital responsibilities enter. Fourteen remain. This is a proposed history for the example, not a claim that every foundation stopped using paper.
Helping an applicant and clarifying missing facts requires listening. A sparse application may need a question, not an assumption about the organization behind it.
The coordinator discusses grant progress, works through change requests and agrees review priorities with the programme team. A programme leader approves a material change; a tidy summary cannot make that decision.
Illustration: five human-led responsibilities receive emphasis. The portal and dashboard responsibilities also remain unchanged in this scenario.
AI can prepare an application brief and draft an applicant reply from permitted records and approved guidance. The coordinator checks the facts and sends the reply; the reviewer judges the application.
AI can also summarize a grantee report or prepare meeting notes. The coordinator checks reported progress against the source, and attendees confirm decisions before the notes become the official record.
Illustration: four preparation tasks become augmented. People accept the output; grant decisions and payments are outside these workflows.
A routine enquiry can reach an internal queue, and an approved document can be indexed against a confirmed grant. AI can interpret wording; ordinary rules enforce identity and access. Ambiguous records return to the coordinator.
A bounded extraction can place stated report figures in a staging table with page references. Checks test units, periods and source matches; uncertain values return to staff. This does not approve a payment or establish the grant’s impact.
Illustration: three internal record routines become automated candidates. Validation and human exception handling remain part of each workflow.
Checking AI summaries becomes a separate responsibility. The coordinator compares the draft with the original submission and records unsupported claims or missing context.
Someone must also maintain the AI checks and examples. When guidance changes or an error repeats, the programme owner accepts the correction before the workflow is reused.
Illustration: two AI-specific responsibilities enter. The final balance is seven unchanged, four augmented, three automated and two new.
Less time preparing briefs and report summaries could leave more attention for applicant questions and progress conversations. That is a possible use of capacity, not a promise that the foundation can handle more grants.
Source checking, corrections and maintaining AI checks take time too. Measure the preparation avoided against that extra work before treating a faster draft as a gain.
Illustration: seven augmented or automated tasks receive visual enlargement. Size is emphasis, not hours saved or a productivity multiplier.
A familiar enquiry may suit a routine more often than a nuanced progress report. Thin evidence, conflicting records or an unusual change request needs staff attention.
The proportions here are illustrative shares of task instances using assistance. They are not measured adoption or time saved. Test the records and checks before extending the workflow to another programme.
Illustration: assumed use splits the seven eligible tasks: briefs 55%, applicant replies 65%, report summaries 40%, meeting notes 45%, enquiry routing 65%, document filing 75%, figure extraction 50%. These are author assumptions, not measured results.
Compare the effort to prepare an accepted brief or report summary with the full effort to check and correct it. A draft that staff cannot use does not count as a useful result.
Include software use, setup, training and maintained AI checks. This planning example claims no savings; repeated accepted work must justify the ongoing effort before the foundation expands it.
Illustration: the remaining tasks and decorative cells form a dollar sign behind four qualitative planning panels. No price, grant value, staffing reduction or payback period is implied.
Measure useful time
Net coordination capacity
Planning measure: subtract checking and upkeep from preparation avoided.
Count the checks
Per accepted result
Planning measure: include rejected attempts and the work to correct them.
Track the workflow
Operating cost
Planning measure: include use, integration and ongoing support.
Test repeat use
Payback
Planning measure: compare recurring benefit with setup and training.
Original MTW scenario informed by primary foundation role descriptions and grant administration sources; no source supplies AI allocations or savings.
AI can draft schedules and compare course material. Instructors assess competence, operations decides when crews are released, and exceptions still need a conversation.
A crew needs time to train, an instructor needs the right material and a learner needs clear instructions. Keeping those responsibilities connected is part of keeping training useful.
AI may help prepare information or organize routine records. A schedule, a course document and a competence decision carry different responsibilities, so they need different boundaries.
A training coordinator plans classes and agrees crew release with operations. They handle last-minute changes, prepare joining details and help learners get support.
This example follows training operations in a large utility. Qualified instructors own instruction, assessors judge competence and curriculum owners approve technical changes. The coordinator keeps the arrangements and records reliable.
Original MTW planning example informed by occupational duties. Earlier methods and AI use are illustrative; no client result or savings is claimed.
In this example, shared calendars and digital enrollment replaced printed timetables, re-keyed registrations and copied rosters. Digital evaluation capture replaced filing paper forms. Those changes happened before AI.
Maintaining course listings and reconciling learner records became ordinary responsibilities. The system still needs the approved course version and evidence for the result it records.
Operations supervisors agree when a crew can attend. The coordinator resolves class changes with the people affected and listens when a learner needs support. A convenient timetable cannot settle those questions alone.
Experts confirm course changes, and assessors resolve disputed training results. A record of attendance does not demonstrate competence; the coordinator seeks the authorized evidence before correcting a record.
AI can propose a class schedule and draft joining details from approved information. The coordinator checks resources and prerequisites, then confirms arrangements with operations and instructors before sending instructions.
AI can compare course revisions and summarize feedback with links to the originals. A curriculum expert checks technical changes; the coordinator checks the feedback summary before an instructor decides how to respond.
A routine query can reach its internal team, and approved course material can receive searchable tags. AI interprets language; ordinary checks enforce the course, revision and audience. Uncertain items return to the coordinator.
Permitted feedback can be grouped into defined topics while keeping the original comments. Unclassified items go to staff; grouping never scores a learner or closes a safety concern. A coordinator still reviews the complete feedback.
Checking an AI course draft means comparing it with the right approved source. The coordinator checks logistics, and a qualified expert checks any technical content before it is used.
Maintaining AI checks becomes recurring work when a course changes. Staff update the examples and stop conditions, then test them so an older accepted answer does not become today’s instruction.
Less schedule preparation and document comparison could leave more attention for class disruptions and learner support. That capacity matters only if the coordinator can use it where the work needs attention.
Checking drafts and maintaining course examples consume time too. Compare the whole workload before treating quicker preparation as greater training capacity.
Established joining details may suit assistance more often than a schedule during an operational disruption. Course versions and reliable learner records limit how far the workflow can be reused.
The proportions here are illustrative shares of task instances using assistance. They are not measured use or time saved. Test another course cycle before expanding to a new training group.
Count the effort to produce an accepted schedule or set of joining instructions, including checking and correction. Expert review belongs in that cost when technical material is involved.
Include integration, software use, setup, staff training and maintained course checks. This example claims no savings or safety improvement; repeated useful work must justify the ongoing effort.
Net coordination capacity: Measure usable time. Planning measure: subtract review and upkeep from preparation avoided.
Per accepted result: Include expert review. Planning measure: count the complete path to accepted preparation.
Operating cost: Include the systems. Planning measure: track workflow use, integration and support.
Payback: Test another cycle. Planning measure: compare repeat benefit with setup and training.
Original MTW scenario informed by O*NET, utility occupational descriptions and relevant training-system experience. No source supplies AI allocations, savings or safety outcomes.
A crew needs time to train, an instructor needs the right material and a learner needs clear instructions. Keeping those responsibilities connected is part of keeping training useful.
AI may help prepare information or organize routine records. A schedule, a course document and a competence decision carry different responsibilities, so they need different boundaries.
Illustration: the broad field represents activities across a utility, not crew numbers, training hours or safety outcomes.
A training coordinator plans classes and agrees crew release with operations. They handle last-minute changes, prepare joining details and help learners get support.
This example follows training operations in a large utility. Qualified instructors own instruction, assessors judge competence and curriculum owners approve technical changes. The coordinator keeps the arrangements and records reliable.
Original MTW planning example informed by occupational duties. Earlier methods and AI use are illustrative; no client result or savings is claimed.
Illustration: sixteen historical tasks form the starting bundle, including coordination, preparation, records and paper routines. This role does not operate the network or independently certify workers.
In this example, shared calendars and digital enrollment replaced printed timetables, re-keyed registrations and copied rosters. Digital evaluation capture replaced filing paper forms. Those changes happened before AI.
Maintaining course listings and reconciling learner records became ordinary responsibilities. The system still needs the approved course version and evidence for the result it records.
Illustration: four historical routines leave during this step and two digital responsibilities enter. Fourteen remain. This is an illustrative transition, not a claim that paper records disappeared across utilities.
Operations supervisors agree when a crew can attend. The coordinator resolves class changes with the people affected and listens when a learner needs support. A convenient timetable cannot settle those questions alone.
Experts confirm course changes, and assessors resolve disputed training results. A record of attendance does not demonstrate competence; the coordinator seeks the authorized evidence before correcting a record.
Illustration: five human-led responsibilities receive emphasis. Course listings and learner-record reconciliation also retain their unchanged category.
AI can propose a class schedule and draft joining details from approved information. The coordinator checks resources and prerequisites, then confirms arrangements with operations and instructors before sending instructions.
AI can compare course revisions and summarize feedback with links to the originals. A curriculum expert checks technical changes; the coordinator checks the feedback summary before an instructor decides how to respond.
Illustration: four preparation tasks become augmented. Schedule suggestions and draft comparisons need acceptance by the responsible people.
A routine query can reach its internal team, and approved course material can receive searchable tags. AI interprets language; ordinary checks enforce the course, revision and audience. Uncertain items return to the coordinator.
Permitted feedback can be grouped into defined topics while keeping the original comments. Unclassified items go to staff; grouping never scores a learner or closes a safety concern. A coordinator still reviews the complete feedback.
Illustration: three internal routines become automated candidates. Grouping comments produces topic buckets; the separate augmented summary interprets the complete feedback with context.
Checking an AI course draft means comparing it with the right approved source. The coordinator checks logistics, and a qualified expert checks any technical content before it is used.
Maintaining AI checks becomes recurring work when a course changes. Staff update the examples and stop conditions, then test them so an older accepted answer does not become today’s instruction.
Illustration: two AI-specific responsibilities enter. The final balance is seven unchanged, four augmented, three automated and two new.
Less schedule preparation and document comparison could leave more attention for class disruptions and learner support. That capacity matters only if the coordinator can use it where the work needs attention.
Checking drafts and maintaining course examples consume time too. Compare the whole workload before treating quicker preparation as greater training capacity.
Illustration: seven augmented or automated tasks receive visual enlargement. The size is emphasis, not a productivity multiplier or a claim about training throughput.
Established joining details may suit assistance more often than a schedule during an operational disruption. Course versions and reliable learner records limit how far the workflow can be reused.
The proportions here are illustrative shares of task instances using assistance. They are not measured use or time saved. Test another course cycle before expanding to a new training group.
Illustration: assumed use splits seven eligible tasks: schedules 40%, joining details 70%, revision comparisons 45%, feedback summaries 50%, query routing 70%, material indexing 75%, topic grouping 60%. These are author assumptions, not measured savings.
Count the effort to produce an accepted schedule or set of joining instructions, including checking and correction. Expert review belongs in that cost when technical material is involved.
Include integration, software use, setup, staff training and maintained course checks. This example claims no savings or safety improvement; repeated useful work must justify the ongoing effort.
Illustration: remaining tasks and decorative cells form a dollar sign behind four qualitative planning panels. No price, safety outcome, headcount reduction or payback period is implied.
Measure usable time
Net coordination capacity
Planning measure: subtract review and upkeep from preparation avoided.
Include expert review
Per accepted result
Planning measure: count the complete path to accepted preparation.
Include the systems
Operating cost
Planning measure: track workflow use, integration and support.
Test another cycle
Payback
Planning measure: compare repeat benefit with setup and training.
Original MTW scenario informed by O*NET, utility occupational descriptions and relevant training-system experience. No source supplies AI allocations, savings or safety outcomes.
People-led work includes reading, review and a decision. The work has value because someone understands what the result needs to achieve.
Identify that decision before changing the process. It provides a standard for judging assistance or automation.
AI-assisted work prepares a draft for human review. The draft is an intermediate result, with a person still deciding whether it is useful.
Keep the review connected to the original task. A fluent draft needs the same attention to facts, context and intent as any other work.
The automated route takes work through validation to a record. It fits repeatable tasks whose requirements can be checked consistently.
Define what happens when validation fails. The process still needs an owner who can investigate and improve it.
People-led, AI-assisted and automated work have different review needs. A team can use all three while keeping responsibility for the result clear.
Look at the tasks inside a role rather than treating the entire role as one thing to automate. Where would assistance make the next decision easier?
People-led work includes reading, review and a decision. The work has value because someone understands what the result needs to achieve.
Identify that decision before changing the process. It provides a standard for judging assistance or automation.
AI-assisted work prepares a draft for human review. The draft is an intermediate result, with a person still deciding whether it is useful.
Keep the review connected to the original task. A fluent draft needs the same attention to facts, context and intent as any other work.
The automated route takes work through validation to a record. It fits repeatable tasks whose requirements can be checked consistently.
Define what happens when validation fails. The process still needs an owner who can investigate and improve it.
People-led, AI-assisted and automated work have different review needs. A team can use all three while keeping responsibility for the result clear.
Look at the tasks inside a role rather than treating the entire role as one thing to automate. Where would assistance make the next decision easier?
Decide how each task reaches an acceptable result: people for judgement, AI to prepare work for review, automation for bounded routines with checks. The cost of a mistake decides which. Plan the new work too: checking sources, maintaining instructions and handling failures.
AI-assisted preparation brings working documents to a reviewer. The purpose is to support a decision with material the person can inspect.
Keep the task and its sources available alongside the draft. The reviewer needs to understand what the work is trying to accomplish.
Human judgement sits between preparation and approved records. The reviewer decides whether the result meets the need and what remains unresolved.
Give that person enough context and time to assess it. Approval is a responsibility in the workflow.
A draft can return for further preparation and review. Specific feedback helps the next attempt address the reason it was not ready.
Keep useful corrections in the working method. That way, the next task can benefit from what the reviewer learned.
Approved records carry the result of the review. Their status tells the next person which work they can use as an accepted starting point.
Preserve enough context to understand the decision. A later change may require another review.
Preparation, review and approval form a working cycle. Returning a draft is part of making the result useful.
The person remains responsible for the decision. Which part of preparation would give them more attention for the judgement only they can provide?
AI-assisted preparation brings working documents to a reviewer. The purpose is to support a decision with material the person can inspect.
Keep the task and its sources available alongside the draft. The reviewer needs to understand what the work is trying to accomplish.
Human judgement sits between preparation and approved records. The reviewer decides whether the result meets the need and what remains unresolved.
Give that person enough context and time to assess it. Approval is a responsibility in the workflow.
A draft can return for further preparation and review. Specific feedback helps the next attempt address the reason it was not ready.
Keep useful corrections in the working method. That way, the next task can benefit from what the reviewer learned.
Approved records carry the result of the review. Their status tells the next person which work they can use as an accepted starting point.
Preserve enough context to understand the decision. A later change may require another review.
Preparation, review and approval form a working cycle. Returning a draft is part of making the result useful.
The person remains responsible for the decision. Which part of preparation would give them more attention for the judgement only they can provide?
An accurate draft can still miss what the client needs. Give the reviewer the sources, context and time to judge it, make it easy to send work back, and keep approved work separate from drafts. Repeated corrections point to a missing rule or a weak example.
A team receives different pieces of work that may look similar at first. Look at their requirements and exceptions before choosing a route.
The useful distinction is what each task needs to reach an acceptable result. Volume alone does not tell you what to automate.
Keep review with people when a decision depends on interpretation or judgement that the process cannot check reliably. They need the context behind the work, not just a finished-looking output.
Make the review responsibility explicit. People need to know what they are deciding and what evidence supports it.
The assisted route uses maintained context to help prepare documents. It changes how the work is produced while leaving the team responsible for its usefulness.
Choose assistance where a person can inspect the result and give actionable feedback. Keep that feedback available for the next attempt.
The automated route follows a defined sequence through validation to a record. Each stage has a clear job in handling the work.
Decide when a failure should stop the sequence and reach a person. Reliable repetition includes a way to handle the exception.
People, assistance and automation can serve different tasks within the same team. Choose the method by the requirements and consequences of the work.
Begin with one repeatable task and judge the result. Broader change can follow what the team learns from doing it.
A team receives different pieces of work that may look similar at first. Look at their requirements and exceptions before choosing a route.
The useful distinction is what each task needs to reach an acceptable result. Volume alone does not tell you what to automate.
Keep review with people when a decision depends on interpretation or judgement that the process cannot check reliably. They need the context behind the work, not just a finished-looking output.
Make the review responsibility explicit. People need to know what they are deciding and what evidence supports it.
The assisted route uses maintained context to help prepare documents. It changes how the work is produced while leaving the team responsible for its usefulness.
Choose assistance where a person can inspect the result and give actionable feedback. Keep that feedback available for the next attempt.
The automated route follows a defined sequence through validation to a record. Each stage has a clear job in handling the work.
Decide when a failure should stop the sequence and reach a person. Reliable repetition includes a way to handle the exception.
People, assistance and automation can serve different tasks within the same team. Choose the method by the requirements and consequences of the work.
Begin with one repeatable task and judge the result. Broader change can follow what the team learns from doing it.
Automate routines with clear inputs and an acceptable result, and route unusual cases, such as a changed recipient or a missing source, to a person. Compare the result with the old process, including review and upkeep, then decide whether to automate more.
A policy, a template and an approved example serve different purposes. Together they explain the rule, the expected form and what accepted work looks like.
Maintain those sources before relying on them for a draft. A better tool cannot settle a contradiction in the underlying guidance.
A SharePoint agent can help answer questions from the sources selected for its scope. Check the answer against the policy and approved example before using it to guide the task.
Keep it distinct from the working draft. The answer informs the work; it does not by itself approve a document.
In this proposed workflow, reusable instructions guide the draft alongside the source material. Configure that method in the chosen agent experience; storing a document in SharePoint does not make it an executable skill.
Check the actual workflow against the team's requirements. Keep the method current when review reveals a missing rule or exception.
A human reviews the draft before it becomes an approved record. The reviewer checks whether the work fits the task and the governing sources.
Carry useful corrections back to the source or method. The next draft should benefit from the review that just happened.
The pattern combines maintained SharePoint sources, an answer, a working method and human review. Each has a distinct purpose on the way to an accepted result.
Start with a document workflow the team knows well. That makes it possible to judge whether the assistance is useful.
Start with the policies, templates and examples your team keeps in SharePoint. An agent answers from content the person can already access, so review permissions first and agree who approves drafts. The SharePoint Online second brain package starts there.
Start with one role and one workflow. Keep what depends on judgement with people, test where AI can prepare work, and decide from the result how far to take it.
Take one role through its recurring work: what stays with people, what AI can prepare, which bounded routines can be automated, and the checking each one needs.
Take one workflow from the map into use, with agreed sources, review built in and a route for exceptions, then compare it with the previous process.
A redesigned piece of work your team can check, run and improve.
Bring a team's recurring work and the exceptions people handle today. Matt helps identify where assistance could be useful, what must stay with people and which first workflow is worth testing.
At enterprise scale, Matt plans and shadows delivery with a larger implementation partner. Smaller scopes can be delivered directly with him.
Start with one role, the decisions it carries and what a mistake would cost.