The five stages of AI adoption, mapped
AI adoption in a professional services firm moves through five stages: Traditional, where most work is still done by hand; Scattered Use, where a few people get results on their own; Built but Fragile, where real workflows depend on whoever built them; Firm Standard, where the method is shared and checked; and Self-Running, where routine steps move on their own and people step in for judgment.
Select a stage on the map to read what it looks like.
Stage 1: Traditional
Most work still happens manually, with little measurable benefit from AI.
Your firm still relies on people completing most of the work by hand. A few people may be experimenting, but occasional experiments don’t move a whole firm past this stage. AI hasn’t yet changed how the business operates.
The work gets delivered. What’s missing is the capacity that routine effort is quietly using up, and that could be recovered.
A normal week
Every firm at this stage looks a little different, but the pattern is remarkably consistent.
Drafting from a blank page
Emails, scopes, reports and client updates are written from scratch, even when last month’s version said almost the same thing.
Filling in the same documents
Engagement letters, intake forms and internal templates are completed field by field, using information the firm already has somewhere else.
Moving information by hand
Someone re-types details from an email into a spreadsheet, then into a system, then into a document.
Meetings written up twice
Notes are taken live, then rewritten afterwards into something a client or colleague can actually read.
Senior people doing junior work
Your highest-paid people still spend hours creating and formatting routine documents that could take a fraction of the time.
None of this is a failure or a broken system. It is simply how the work has always been done, and it is costing you time and money.
At a glance
Exit test: One person completes the pilot task faster than its recorded baseline while meeting an agreed quality standard.
- Where you are
- Work is still mostly manual. AI use is limited, informal or nonexistent.
- What it costs
- Time disappears into drafting, re-entry, coordination, review and repetitive admin.
- The constraint
- Where manual effort absorbs most of the day, it caps capacity, and hiring becomes the only visible way to add more.
- What changes next
- Start proving AI in a few measurable workflows.
Inside the Stage 1 roadmap
Stage overview 
What it looks like 
At a glance
Not sure this is your stage? Find yours in two minutes.
Stage 2: Scattered Use
Your team is using AI, but the results depend on who's using it.
AI is starting to show up in real work. A few people are seeing early wins, but adoption is still uneven across the firm.
The wins are happening. What’s missing is making them consistent, shared and repeatable.
A normal week
The tools are in the building, and a few people are getting real value from them.
A few people are much faster
One or two employees have figured out how to use AI well and are saving real time. Most of the team still works the old way.
Most use stops at quick questions
Reusable instructions, reference documents, structured outputs, deeper research and approved connections to work information go untouched.
Useful setup gets rebuilt repeatedly
People rebuild prompts, client context, tone and formatting every time instead of using reusable workflows, templates or shared standards.
The best use cases stay with one person
When someone finds a better way to work, it usually lives in their own account or in their head. The rest of the team never benefits.
Nobody is completely sure what is allowed
People hesitate, self-censor or use tools quietly, because the firm hasn’t defined which tools are approved, what data is safe, or how AI fits into client work.
The tools are already there. The gap is making the best use cases repeatable. The test: can three people complete the same task to the same standard?
At a glance
Exit test: The workflow repeatedly hits its time and quality targets in real work.
- Where you are
- AI tools are in place, but usage is uneven and concentrated in a few people.
- What it costs
- Time, money and capacity are lost to inconsistent output, rebuilt work and underused licenses.
- The constraint
- The firm has the tools, but no shared system for turning individual wins into repeatable workflows.
- What changes next
- Package the best use case with a written output standard, and make it repeatable within a defined pilot group.
Inside the Stage 2 roadmap
Stage overview 
What it looks like 
At a glance
Not sure this is your stage? Find yours in two minutes.
Stage 3: Built but Fragile
Your AI workflows are delivering value, but keeping them working still depends on key people.
AI is creating real value in parts of the firm. But the results still depend on specific people, prompts and processes.
The wins are real. What’s missing is making them durable enough to survive a vacation, and structured enough to scale.
A normal week
AI is already saving time in real workflows. The problem is that those wins still depend on the people who figured them out.
A few workflows are working really well
Certain recurring tasks are noticeably faster, and the output holds up in real client work.
It runs on one person and their accounts
The builder holds the prompts and the exceptions, and parts of it may sit in a personal account or subscription the firm doesn’t control.
The instructions have gone out of date
What is written down is dated and no longer matches the firm’s current standards.
Unusual cases are handled inconsistently
The straightforward cases go smoothly. The awkward ones get judgment calls that differ by person, with no agreed rule for what to do.
It breaks when the source material changes
When a template, policy or reference document is updated, the workflow keeps using the old version until someone notices and repairs it by hand.
The workflow produces value, but the firm struggles to maintain it, handle unusual cases, or extend it without its builder.
At a glance
Readiness condition: The required access works on firm accounts, a backup owner is trained, exception cases are covered, and results stay acceptable when the builder isn’t involved.
- Where you are
- Real workflows are working and saving time, but they depend on the people who built them.
- What it costs
- Results swing with availability, and scaling a workflow means rebuilding it from scratch.
- The constraint
- Continuity and support. The method works, but nobody else can keep it running or repair it.
- What changes next
- Prove the relevant team can use and support the approved workflow without relying on its builder.
Inside the Stage 3 roadmap
Stage overview 
What it looks like 
At a glance
Not sure this is your stage? Find yours in two minutes.
Stage 4: Firm Standard
Your team has a shared method. The next opportunity is reducing routine handoffs.
The firm has adopted a shared, supported way of working: documented, taught and expected for the work and roles it applies to.
Routine delivery no longer depends on one particular person. What’s missing is the next step: the work still needs people to push it from each step to the next.
A normal week
The better way of working is no longer personal. It is documented, taught and built into how the firm operates.
The workflow is written down
The steps, prompts, inputs, outputs and exceptions live in approved business accounts and controlled shared material that authorized staff can follow, not in personal ones.
New people learn it in onboarding
A new hire is taught the firm’s method from day one, instead of spending months working out how someone else does it.
Quality is checked, not hoped for
There is a clear definition of good output, what needs review and who signs off. Output still varies; what changes is that it is checked before it goes out.
It keeps running when people are away
Vacation, illness or turnover doesn’t stop the workflow. It is supported inside the systems staff already use, and the knowledge belongs to the firm.
Improvements are made once and shared
One person owns the approved version and reviews the material it depends on. Out-of-date reference documents get caught rather than quietly used.
The firm now owns the method. The next opportunity is to remove the manual steps that still depend on someone remembering to act.
At a glance
Readiness condition: The approved workflow runs across the relevant roles with output consistently checked, and the remaining cost is the manual coordination between steps, not the work itself.
- Where you are
- The approved workflow is documented, taught, supported and used across the roles it applies to.
- What it costs
- People still start each run, chase status, route exceptions and confirm the next step happened.
- The constraint
- Coordination. The method is solid, but nothing moves until someone notices and pushes it forward.
- What changes next
- Remove the manual coordination between steps, keeping judgment, exceptions and oversight with people.
Inside the Stage 4 roadmap
Stage overview 
What it looks like 
At a glance
Not sure this is your stage? Find yours in two minutes.
Stage 5: Self-Running
Selected workflows run automatically. The priority is keeping them reliable and choosing what to improve next.
Selected workflows begin moving on their own. The right people step in only for judgment, approvals and exceptions.
The firm no longer depends on someone to keep routine work moving, so speed, consistency and scale improve together.
A normal week
Selected workflows now start, route and advance on their own. People step in for judgment, approvals and exceptions, not routine coordination.
Work starts on its own
When the trigger conditions are met, the workflow begins. Nobody has to remember to start it or notice that it is ready.
Routing happens automatically
Work moves to the right person or system with its context attached.
People step in for exceptions
The team is notified when something needs judgment or approval, or falls outside the standard path.
Cycle time drops, not just task time
Once the waiting between steps is gone, the whole process finishes far faster, even when the work itself hasn’t changed.
Reliability is monitored, not assumed
Every workflow has an owner, alerts and a fallback, and the review asks more than whether it ran: was the output right, did the correct record update, did it stay within cost limits?
One connected example: a signed engagement letter lands in the shared drive, an agent drafts the onboarding summary from approved reference documents and writes it to the client record, and the engagement manager approves it before anything reaches the client. A missing detail pauses the run and alerts its owner.
Keeping the gains
As more routine work runs automatically, a few simple habits keep the benefits lasting.
- Make sure someone else can step in
- Give someone responsibility for the workflow, make sure a backup knows how to run it, and keep instructions and access available to the people who need them.
- Know when something needs attention
- Notify the right person when work gets stuck or needs a decision, and keep a simple way to finish the task by hand.
- Check that it’s still helping
- Regularly review the results, time saved and running costs, and update the workflow when the business changes.
- Improve one thing at a time
- Choose the next repetitive task, get the process working consistently, then automate the steps that make sense.
You don’t need to automate everything. Start with one useful improvement, prove it works, and build from there.
Inside the Stage 5 roadmap
Stage overview 
What it looks like 
Keeping the gains
Not sure this is your stage? Find yours in two minutes.
Common questions
Questions about the five stages
Traditional, Scattered Use, Built but Fragile, Firm Standard and Self-Running. Each stage describes how the work actually gets done rather than which tools the firm pays for: whether AI is used at all, whether results depend on a few individuals, whether a workflow survives without the person who built it, and whether routine steps move without someone pushing them.
Think about normal day-to-day work, not your most advanced experiment. A firm where one person has built something impressive but everyone else works the old way is at Stage 2 or 3, not Stage 5. The free assessment at sentinelconsulting.ai/roadmap places your firm on the map in about two minutes.
Each roadmap explains the stage, shows what a normal week looks like inside it, puts a number on what the stage costs a firm your size in hours and dollars, and lays out a plan for reaching the next stage, with a one-page summary to share with your team. The Stage 5 roadmap covers keeping automated workflows reliable instead.
The roadmaps are written in six editions, from a solo practice to firms of more than 100 people, because what a stage costs depends on how many people are doing the work. Your firm size picks the edition with the right numbers. The email is so one of the founders can follow up personally; there is no newsletter.
Find where AI pays in your business.
The audit is free and the roadmap is yours to keep, whether you build it with us or not.
Free · No pitch · A plan you keep