AI for professional services
AI for professional services firms
Proposals, status reporting, and the firm's collective knowledge, working on every engagement. 5x ROI in 30 days, or we work free.
- Consultancies
- Agencies
- Engineering firms
- MSPs
- Dev shops & integrators
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Senior time goes to non-billable work.
Proposals, status decks and internal reporting, done by the people whose hours clients actually buy.
Every engagement starts from zero.
The firm solved this exact problem two years ago. Nobody can find where.
Utilization and scope creep hide until month-end.
By the time the numbers land, the margin is already gone.
Scope creep gets billed as goodwill.
The tenth revision ships free because nobody raised the change order in week two. The retainer absorbs it, and the margin review finds it in ninety days.
The method leaves with the senior who resigns.
The scoping logic, the client history, the reason the last engagement worked. Four weeks of notice against ten years of accumulated judgement.
The RFP deadline sets the thinking time.
The response is due Friday, so the blank page gets partner hours the client will never see on an invoice, and the scoping is done at midnight speed.
Professional services, before and after
The manual path is dashed: Partners writing proposals, Past work unfindable, Margin news arrives late. The system path replaces it, and a person approves before anything ships: Drafts from your best work, Firm knowledge on tap, Utilization visible weekly.
Before: by hand
- 01Partners writing proposalshuman
- 02Past work unfindablehuman
- 03Margin news arrives latehuman
After: the system
- 01Drafts from your best work
- 02Firm knowledge on tap
- 03Your approvalhuman
- 04Utilization visible weekly
The research
Small-business AI adoption reached 63-88% depending on survey, with productivity the #1 motivation (BizBuySell, 2026).
BizBuySell · 2026
Proof
An Ontario professional-services firm runs a knowledge brain over its own published material today. The case below reports adoption and usage read from the tenant database. It does not report hours back yet, and neither does this page.
We guarantee 5x ROI inside 30 days of deployment, in writing, measured against a baseline you sign before we build. If the system misses the bar, we keep working for free until it clears.
The numbers in professional services firms
Small-business AI adoption reached 63-88% depending on survey, with productivity the #1 motivation (BizBuySell, 2026).
BizBuySell · 2026
95% of GenAI pilots show no measurable P&L return. The winners pick use cases by P&L impact, not novelty (MIT, 2025).
MIT NANDA · 2025 · 150 interviews, 350-employee survey, 300 deployments analyzed
A Tuesday running three engagements
Overnight status assembly.
Timesheets from Harvest, task movement from Asana and yesterday's client emails are pulled into one status draft per engagement, each line cited to its source record.
Morning review and release.
humanThe engagement manager reads three drafts in ten minutes, corrects the milestone date on one, and releases them. Nothing reaches a client until this click. This is an approval gate, not a formality.
Burn check across the portfolio.
In this morning's run, one engagement has consumed 60 percent of budgeted hours at 40 percent of milestones. The system flags it and drafts a change-order note with the timesheet evidence attached.
Change-order decision.
humanThe engagement manager takes the flag to the partner. Raise the change order, absorb the overrun, or rescope. The system never makes this call and never contacts the client about money.
Meeting-to-action capture.
Notes and decisions from the morning's kickoff call are filed to the engagement record, and the six action items are routed to their owners in Asana with due dates.
Proposal skeleton for the new RFP.
An RFP landed at noon. The system retrieves the two closest past engagements from the deliverable archive and assembles a first-pass proposal with the win themes and the old scoping logic, sources cited.
Partner shapes the proposal.
humanThe partner spends the afternoon hour on positioning and price instead of the blank page. The draft leaves the building under a person's name after a person has rewritten what matters.
Who this is built for
Sunday goes back to being Sunday.
The draft is assembled from the firm's past wins before you open the file. Your job shrinks to the part clients actually pay a partner for, the scoping judgement and the price. We publish no hours-back number yet. The professional services case we have published measured adoption and usage, not hours, and your engagement sets that number on your own timesheets against a baseline you sign.
Status day becomes a ten-minute review.
Every engagement's status assembles overnight from the timesheets in Harvest and the tasks in Asana. You read three drafts, fix one number, and release. The half day you used to spend building the deck goes back on the engagement it was billed against.
Margin news arrives in week two, not month-end.
Burn against budget per engagement, every Monday, from the timesheet data you already collect. When hours outrun milestones, the change-order draft is waiting with the evidence attached. Scope creep stops being a write-off you discover at invoicing.
The last engagement briefs the next one.
Ask what the firm did the last time a client had this exact problem and get the deliverable, the decision memo and who worked on it, with the file cited. You start from the firm's best prior answer instead of a blank page, and you can check the source before you trust it.
What stays human
- Morning review and release
- Change-order decision
- Partner shapes the proposal
The steps the day below leaves to a person, by design.
A Tuesday running three engagements
Managing partner runs the day through the built system: Overnight status assembly, Morning review and release, Burn check across the portfolio, Change-order decision, Meeting-to-action capture, Proposal skeleton for the new RFP, Partner shapes the proposal. Dashed steps stay with a person.
- 01Managing partnerhuman
- 02Overnight status assembly
- 03Morning review and releasehuman
- 04Burn check across the portfolio
- 05Change-order decisionhuman
- 06Meeting-to-action capture
- 07Proposal skeleton for the new RFP
- 08Partner shapes the proposalhuman
What we build for professional services firms
Win the work
Proposal and SOW automation.
Drafted from your past wins and your scoping logic. A partner reviews instead of writes.
Client onboarding automation.
New engagements papered, provisioned and kicked off without partner chasing.
Run the engagement
Engagement status reporting.
Clients and partners see where things stand without anyone assembling a deck.
Meeting-to-action pipeline.
Notes, decisions and tasks captured from every client meeting and routed to the people who own them.
Utilization and margin reporting.
Scope creep and utilization visible weekly, while there is still time to act.
Change-order drafting.
When burn outruns milestones, the draft is waiting with the timesheet evidence attached. Raising it stays a partner's call.
Keep what the firm learns
Firm knowledge brain.
Every deliverable, decision and lesson the firm has produced, searchable, so engagements stop re-learning.
Close-out capture.
At engagement close, the final deliverables, the decisions and the win-loss notes file themselves into the archive. The brain grows without anyone playing librarian, and the method survives the resignation.
The economics
Before and after economics
Line
- Active use of the AI tools the firm already licensed
Before
Licences bought, a few people using them, nobody measuring
After
Usage measured per person, because the system is built into the work people already do
- Hours returned per person, per week
Before
Unknown, because nobody recorded the baseline
After
Measured on your own timesheets against a baseline you sign, and the 5x ROI guarantee is judged on it
- Non-billable assembly across a 7-person delivery team
Before
63 hours a week, roughly $6,930 at loaded cost
After
Assembly runs on a schedule. People keep the review, and the hours move back to billable work
No row on this table is a client outcome. The professional services case we have published reports adoption and usage counts, not hours or dollars, and this page does not borrow a number the case did not measure. The team row multiplies this page's calculator defaults, 7 people spending 9 hours a week each on proposals, status and chasing, valued at the $110 default loaded hourly cost, which is 63 hours or $6,930 a week in scope. Every other row is a qualitative state change. Your numbers replace the defaults in the calculator below.
Where the data comes from
Project mgmt
Whether the firm runs Asana, ClickUp, Teamwork or ConnectWise, this is where the engagement actually lives: task status, milestones, budget burn and the timesheets behind utilization. Status reporting and the weekly margin view read from here, and meeting actions are routed back into it.
Where it stops. Read in place. The only writes are tasks a person routed, and one engagement's burn numbers never render in another client's context.
CRM
HubSpot, Pipedrive or Salesforce holds the pipeline, the contacts and the win-loss history that proposal drafting draws its themes from. Onboarding automation starts here the day the SOW is signed.
Where it stops. Prospect data stays on the business development side of the wall. It never appears in client deliverables, and no contact is messaged without CASL consent state checked in the data layer first.
Deliverable archive
The SharePoint, Google Drive or Notion estate where every proposal, final report and decision memo the firm has shipped ends up. Indexed, it becomes the corpus: the reason the next engagement starts from the firm's best prior answer, with the source file cited.
Where it stops. Indexed under per-client scoping, so a query from one engagement can only retrieve what its scope allows. Nothing leaves the firm's own accounts and no model trains on any of it.
How the system is built for professional services firms
See the full capability mapRetrieval
Every proposal, deliverable and project record the firm has produced, indexed so the next engagement starts from what the last one learned. Answers cite the file, so a junior can check a senior's work.
- pgvector
- Full-text BM25
- Reciprocal rank fusion
Agents and orchestration
Agents draft proposals, assemble status reporting and keep the pipeline current on a schedule. Client-facing output is a proposal a person releases. The judgement stays billable and human.
- agent-worker
- Scheduled runs
- Proposal queue
Evaluation
Graded against work the firm has already delivered and stood behind. Retrieval accuracy and citation correctness are tracked over time, so quality is a trend line rather than an impression.
- Eval graders
- Citation grading
- quality-worker
Models
Frontier models for drafting and synthesis across long project records. Embedding models for retrieval across the archive. Nothing exotic where a simpler method does the job.
- Frontier models, one gateway, routed per task
- Embedding models via the same gateway
- Rules where they suffice
Data boundary
Per-client scoping at the database layer, which is how client confidentiality survives contact with a shared knowledge base. Nothing crosses engagements by construction.
- Supabase row-level security
- Per-client scoping
- No-training API terms
Project mgmt, CRM, Deliverable archive feed a hybrid index. The agent runtime works from that index, and every consequential action passes a human approval before it reaches Ask your brain, Knowledge map, Workflow builder.
Your systems
- 01Project mgmt
- 02CRM
- 03Deliverable archive
The system
- 01Hybrid index
- 02Agent runtime
- 03Your approvalhuman
Where your team works
- 01Ask your brain
- 02Knowledge map
- 03Workflow builder
What that means in practice
Retrieval and RAG
Connecting AI to your actual documents so it answers from your knowledge, accurately and with citations, instead of making things up.
Where we stop. A dedicated vector database is justified by scale, not by default. Most of RAG quality is won or lost in chunking and indexing strategy, not in the model choice, and we have walked clients back from RAG to plain search when that was the honest answer.
Agents and orchestration
AI that does the work instead of just answering: looks things up, calls your systems, completes multi-step tasks, and knows when to hand off to a human.
Where we stop. Multi-agent swarms are oversold; most jobs need one well-guarded loop. If a cron job and a script solve it, that is what we build, because 90 percent per-step accuracy compounds to 59 percent over five chained steps and no framework changes that arithmetic.
Evaluation and observability
How we prove the AI actually works: measured, monitored and regression-tested like real software, not vibes.
Where we stop. There is no engagement where we skip this. The honest variable is depth: a document pipeline gets faithfulness and extraction suites, an outbound agent gets human review sampling, a classifier gets a held-out test set. We size the harness to the risk, never to zero.
Built around your rules
| Regime | What it demands here | How the system complies |
|---|---|---|
| Client confidentiality and NDAs | One client's work must be invisible to every other client, and the firm must be able to prove it, not just promise it. | Per-client scoping at the database layer with row-level security, access logged. The NDA is enforced in the architecture, not just the contract. |
| Professional standards (law societies, CPA bodies, provincial engineering regulators) | Regulated deliverables go out under a member's name, seal or signature, and the regulator holds that person responsible regardless of what drafted the file. | Outputs stay drafts a licensed professional reviews and signs. An engineering drawing is sealed by the P.Eng who reviewed it, never released by a system, and the firm's project records stay in the tools the regulator would expect to audit, whether that is Deltek Vantagepoint or BQE Core. |
| PIPEDA and BC PIPA | Personal information collected in the course of engagements is handled with consent, purpose limits and access rights under federal and BC privacy law. | The personal information in play is engagement personnel data, the names, roles and contact details of client staff gathered in the course of the work. It stays in the firm's accounts, handled to respect both statutes, with residency scoped per engagement and access logged and reviewable. |
| CASL | Commercial electronic messages require consent, sender identification and a working unsubscribe. | Business development outreach is consent-based, identified and unsubscribable, with consent state enforced in the data layer so an agent cannot message a contact who opted out. |
| Client security reviews and SOC 2 questionnaires | Enterprise clients audit their vendors, and MSPs and consultancies inherit those questionnaires whether or not they hold the attestation themselves. | Data stays in the firm's own accounts under no-training API terms with an audit trail, so the questionnaire answers are true as written rather than aspirational. |
The regimes that govern professional services, what each demands, and how the system complies
Regime
- Client confidentiality and NDAs
What it demands here
One client's work must be invisible to every other client, and the firm must be able to prove it, not just promise it.
How the system complies
Per-client scoping at the database layer with row-level security, access logged. The NDA is enforced in the architecture, not just the contract.
- Professional standards (law societies, CPA bodies, provincial engineering regulators)
What it demands here
Regulated deliverables go out under a member's name, seal or signature, and the regulator holds that person responsible regardless of what drafted the file.
How the system complies
Outputs stay drafts a licensed professional reviews and signs. An engineering drawing is sealed by the P.Eng who reviewed it, never released by a system, and the firm's project records stay in the tools the regulator would expect to audit, whether that is Deltek Vantagepoint or BQE Core.
- PIPEDA and BC PIPA
What it demands here
Personal information collected in the course of engagements is handled with consent, purpose limits and access rights under federal and BC privacy law.
How the system complies
The personal information in play is engagement personnel data, the names, roles and contact details of client staff gathered in the course of the work. It stays in the firm's accounts, handled to respect both statutes, with residency scoped per engagement and access logged and reviewable.
- CASL
What it demands here
Commercial electronic messages require consent, sender identification and a working unsubscribe.
How the system complies
Business development outreach is consent-based, identified and unsubscribable, with consent state enforced in the data layer so an agent cannot message a contact who opted out.
- Client security reviews and SOC 2 questionnaires
What it demands here
Enterprise clients audit their vendors, and MSPs and consultancies inherit those questionnaires whether or not they hold the attestation themselves.
How the system complies
Data stays in the firm's own accounts under no-training API terms with an audit trail, so the questionnaire answers are true as written rather than aspirational.
Client engagement data is scoped per account with row-level security, under no-training API terms, with an audit trail your clients can ask about.
The objections
Our work is bespoke. No two engagements are alike.
The deliverable is bespoke. The assembly around it is not. Proposals follow your scoping logic, status follows your timesheets, onboarding follows your SOW. We automate the repeated shape and leave the judgement, which is the part clients are buying anyway.
Partners will never trust a drafted proposal.
They should not trust it blind, and the system does not ask them to. Every draft cites the past engagement it drew from, so the partner checks sources instead of taking dictation. Reviewing a cited draft is faster than writing, and the partner's name only goes on what the partner rewrote.
We built a knowledge base in SharePoint once. It died.
It died because filing was a chore and search returned folders. Here nothing depends on people filing: the archive indexes automatically at engagement close, answers cite the file they came from, and retrieval accuracy is graded over time rather than assumed.
Will clients push back on AI touching their engagement?
Some will ask, and you will have a real answer. Their data is scoped to their engagement at the database layer, under no-training API terms, with an access log you can show them. That answer is more than most firms can say about their file shares.
Weekly utilization reporting will feel like surveillance.
It reads the same Harvest timesheets you already collect and reports at the engagement level, where the margin question actually lives. The change from month-end to weekly is about catching scope creep while the change order can still be raised, not about watching people.
How we build it for professional services firms
Index the deliverables you already shipped.
The firm's own proposals and project records become the corpus, so the next engagement starts from what the last one learned instead of a blank page.
Automate the assembly, keep the judgement billable.
Proposals and status reporting assemble from real project data. The thinking stays with the people clients are paying for.
Extend to pipeline once client scoping holds.
Per-client scoping is tested before a shared knowledge base touches a second client. That is the thing that ends the engagement if it is wrong.
What we will not automate
Client advice and anything sent under the firm's name without a person releasing it.
Where your team works
Tour the platformAsk your brain
Ask a question. Get the answer and where it came from.
Every answer cites the document behind it, so you can check the source yourself. When the brain does not have the answer, it says so instead of guessing.
Knowledge map
See everything the brain knows on one map.
Every document, person, and decision the brain has learned, drawn as one connected map. Thin spots are visible too, so you know exactly what to feed it next.
Workflow builder
Describe the workflow. Watch it assemble.
Say what should happen in plain language and the builder assembles the automation on a canvas you can read, run, and change.
Goals
Goals that report their own health.
Each objective shows the work feeding it, how healthy it is today, and where it is projected to finish. You find out early, not at the deadline.
By team
The same system, seen from the desk that runs it.
Run your numbers.
Your operations
Savings use the low end of our hours-reclaimed range. The math is conservative on purpose.
The math
Calculated at the low end of every range.
Every first build is covered in writing: 5x ROI in 30 days. Or we work for free.
The hard questions
The systems behind this
Free · 3-5 days
Know your number in five days.
We map your operations, find the highest-ROI automations, and hand you a ranked plan with the payback math attached. Yours to keep, whoever builds it.
Prefer to talk first? Book 15 minutes with James. No pitch deck.