Industry Guide Consultancy
Individual AI productivity rarely becomes firm-level margin on its own. Map client delivery workflows, redesign review and pricing around AI, and measure results against a baseline.
Most accounting, legal, consulting and marketing firms now use AI somewhere. Fewer can show what it has done for the firm. Thomson Reuters' 2026 AI in Professional Services report, drawing on more than 1,500 professionals globally and published in April 2026, found organization-wide AI adoption had risen from 22% to 40%, yet only 18% of firms were tracking AI's business value.
That gap is the difference between AI adoption and AI transformation. Adoption gives individuals faster tools. Transformation changes how client work flows through the firm, who reviews it, how it is priced and how results are measured. McKinsey's State of AI 2026 points in the same direction across industries: nearly three-quarters of the organizations it classed as AI high performers reported fundamentally redesigning workflows around AI, compared with about a quarter of other respondents.
This guide sets out a practical approach for professional services firms: map delivery workflows, prioritize measurable use cases, redesign review roles and quality gates, adjust the operating model, and measure outcomes against a baseline.
In a firm that bills by the hour, an associate who drafts a memo in half the time does not automatically improve the firm's results. The saved time may be absorbed by other work, written off, or turned into lower bills. Meanwhile, partners may spend more time checking AI-assisted work they do not trust. Without changes to workflow, review and pricing, AI can make individuals faster while leaving margins, turnaround times and client experience unchanged.
Transformation therefore needs to answer three questions together: which steps of client delivery change, who is accountable for quality at each step, and how the firm captures the value.
Pick one or two service lines that matter commercially, and map how a typical engagement moves from start to finish. A generic professional services workflow often looks like this:
| Stage | Typical activities | Who does it today |
|---|---|---|
| Intake and scoping | Understand the request, check conflicts or independence, agree scope and fees | Partner or manager |
| Information gathering | Collect documents and data from the client | Junior staff and client contacts |
| Research and analysis | Research rules, precedents, markets or data; analyze | Associates and seniors |
| Drafting | Prepare memos, reports, filings, models or creative work | Associates and seniors |
| Review | Check accuracy, judgment, compliance and quality | Managers and partners |
| Delivery and advice | Present findings and recommendations | Partner or manager |
| Billing and close-out | Record time, invoice, archive | Staff and finance |
For each stage, record a baseline using recent engagements: elapsed time, hours by role, rework loops, write-offs and common client complaints. That baseline is what you will measure against later. Without it, any claimed improvement is anecdote.
List the places where AI could change the work, then score them. A simple approach uses four factors from 1 to 5:
Examples by profession, for illustration only:
| Profession | Candidate use case | Notes |
|---|---|---|
| Law | First-pass review of contracts against a playbook | High volume; human review of every flagged issue |
| Accounting | Organizing and extracting data from client documents for preparation | Clear inputs; outputs checked against source |
| Consulting | Research synthesis for market assessments | Requires source verification and analyst judgment |
| Marketing agencies | First drafts of routine content variants | Brand and factual review essential |
| All | Meeting notes, action tracking and status updates | Low risk; quick to adopt |
When choosing the first service line, look for one with enough volume to measure results within a few months, a practice leader willing to sponsor change, and clients who will not object to new ways of working. A service with steady, repeatable engagements, such as recurring compliance work or standard assessments, usually produces clearer evidence than bespoke, one-off advisory work.
Favor use cases that are high value, feasible and measurable, and where the review step can be defined clearly. Postpone use cases where an undetected error would be costly until review standards are proven elsewhere.
This is where transformation succeeds or fails. When AI produces first drafts, the critical skill shifts from producing to verifying and judging. Redesign deliberately:
Define quality gates. For each redesigned workflow, write down what must be checked before work moves to the next stage, for example that every citation has been verified against the source, every number reconciles to client data, and every recommendation is supported by evidence in the file.
Assign accountable reviewers. Make clear who signs off at each gate. Professional responsibility does not transfer to a tool. The American Bar Association's Formal Opinion 512, for example, reminds lawyers using generative AI of their duties of competence, confidentiality, client communication and reasonable fees.
Keep an audit trail. Record which steps used AI, which sources were used and who reviewed the output. This supports quality control, client questions and regulatory inquiries.
Protect learning. Junior professionals learn by doing first drafts and research. If AI does more of that work, redesign how juniors build judgment, for example by having them critique and correct AI drafts with a reviewer explaining the corrections. Our guide to AI upskilling for professional services teams covers the training side.
Clarify client communication. Decide how and when clients are told about AI use, and whether engagement terms need updating.
Workflow changes need supporting changes in how the firm runs:
Measure the redesigned workflow against the baseline from step 1, over enough engagements to be meaningful:
| Measure | What it shows |
|---|---|
| Cycle time from intake to delivery | Client-visible speed |
| Hours by role per engagement | Change in effort and leverage |
| Realization and write-offs | Whether efficiency turns into margin |
| Review findings and rework | Quality effect |
| Client satisfaction and repeat work | Client-perceived value |
| Margin per engagement or service line | Firm-level financial result |
| Adoption of the redesigned workflow | Whether the change has actually happened |
Compare like with like: engagements of similar size, complexity and client type, run before and after the change. Where volumes are small, combine quantitative measures with structured reviewer feedback rather than drawing conclusions from two or three engagements.
Report results to partners in business terms, not tool usage statistics. A finding such as "market assessments now take eight working days instead of fourteen, with no increase in review findings, and margin per assessment rose" is far more persuasive than logins per month.
The following example is invented for illustration.
A consultancy delivers market assessments that typically take three weeks and involve a team of three. Mapping shows that roughly half of the hours go on research gathering and synthesis, and that partners rewrite a large share of draft sections because sources are unclear.
The firm redesigns the workflow:
It measures cycle time, hours by role, review findings and margin for the next ten assessments against the previous ten, and decides whether to extend the approach to due diligence work.
Partner skepticism. Senior professionals often distrust AI output because they have seen confident errors. Involve them in designing the quality gates, and show them reviewed results from real engagements rather than demonstrations.
The billable-hour conflict. If a practice is measured on hours billed, faster work looks like lost revenue. Address pricing and incentives early, even if only for one service line as a test.
Scattered experiments. Many firms have dozens of individual experiments and no shared standards. Consolidate them into a short list of priority workflows with owners, and retire the rest or leave them as personal productivity aids.
Knowledge that is hard to find. AI is most useful when it can draw on the firm's own precedents, templates and past work. If documents are scattered across personal drives and inconsistent folders, fixing document management may be the first real step.
Client concerns. Some clients restrict AI use in their engagements. Track client-specific restrictions centrally so teams know where redesigned workflows may and may not be used.
| Period | Focus |
|---|---|
| Days 1 to 30 | Choose one or two service lines; map workflows; capture baseline from recent engagements; confirm approved tools and data rules |
| Days 31 to 60 | Redesign the workflow with practice leaders; define quality gates and reviewer roles; train the pilot team; decide pricing for the pilot |
| Days 61 to 90 | Run the redesigned workflow on live engagements; measure against baseline; review findings with partners; decide whether to extend, adjust or stop |
Ninety days is enough to produce evidence from real client work. It is not enough to transform a firm, but it gives leadership a defensible basis for the next investment decision.
AI transformation is changing how a business delivers its work and makes money with AI: redesigning workflows, review roles, pricing and governance so that faster individual tasks turn into better margins, shorter cycle times or higher quality for the firm, not just for individual users.
AI adoption gives people faster tools. AI transformation changes the work itself: which steps of client delivery change, who is accountable for checking AI-assisted output, and how the firm prices, staffs and measures engagements so the time saved shows up in results.
Capture a baseline from recent engagements first: elapsed time, hours by role, rework, write-offs and margin. Then compare the redesigned workflow against that baseline over enough similar engagements, and report in business terms such as margin and cycle time, not tool usage.
Choose one service line, map its delivery workflow and capture a baseline from recent engagements. That exercise alone usually reveals two or three high-value use cases and the review changes they require. For the systems side of the firm, such as connecting CRM, billing and onboarding, see our guide to client onboarding automation for professional services firms.
Growix's Digital Transformation team runs technology-enabled change as a business program, covering strategy, process, technology and capability, and our Operating Models & Business Process Optimisation team redesigns workflows, governance and decision rights.
Digitizing a broken process makes it fail faster. Diagnose whether your constraint is process or technology, compare scope and disruption, and sequence the work with milestones.
An assessment tells you where AI fits and what blocks it; a pilot proves one use case works. Use this decision tree to pick the right first step, or combine them.
When one location wins customers and another loses them, the difference is usually in the journey, not the service itself. Map it with real data, compare sites and standardize what matters.
