A professional services firm sells time and judgment. Consultancies, agencies, brokers, accountants, advisors and B2B service companies all have the same hidden cost: the hours between the judgment, spent sorting inquiries, writing proposals from scratch, chasing invoices and typing up meetings.
AI in professional services firms is mostly about recovering those hours. The uses below are working in firms we have built for. A few are not ready yet, and the difference is easy to tell once you know what to look for.
The productivity gap between people and profit
McKinsey's 2026 survey found that 80% of respondents said AI improved their own productivity, but only 37% saw any effect on earnings. Services firms feel this acutely. Everyone has a chatbot open; nobody can point to the invoice it shortened or the deal it won.
The firms closing that gap, in our research on AI adoption in mid-size businesses and in our own projects, do one thing differently. They put AI inside a specific workflow with a measurable output, rather than giving staff a general tool and hoping.
Six workflows where it is paying back
1. Inquiries sorted before anyone reads them
For a B2B firm we connected the shared inbox to the CRM so every incoming inquiry is read, classified, matched to a company and logged with a summary and a suggested owner. Spam and vendors go one way, real prospects go to the right person with context. The firm's response time dropped because nobody was triaging email by hand first thing in the morning.
2. Proposals drafted from your own past bids
Firms that respond to tenders answer the same questions many times. We built an assistant that drafts responses to a new request for proposal from the firm's library of past bids, pulling the paragraphs that answered similar questions and flagging where the new request differs. The bid team edits instead of starting from blank, and the firm's best answers stop living in one person's folder.
3. Follow-ups that happen on time, in the firm's voice
A sales follow-up agent reads the CRM, notices which conversations have gone quiet, and drafts the next email in the style of the account owner, who approves or edits before anything is sent. Deals stop dying of neglect. The approval step is not optional; it is what makes the firm comfortable letting the agent draft at all.
4. Meetings that update the CRM themselves
Call notes rarely make it into the system. An assistant that summarizes each meeting, extracts the actions and creates the CRM tasks, for the account owner to confirm, means the pipeline reflects what was actually said. Combined with a daily digest in Slack or Teams of open tickets, pipeline changes and overdue tasks, managers stop asking for status updates.
5. Invoices chased politely and consistently
Receivables are where services firms quietly lose cash. An accounts receivable agent we built tracks which invoices are due, sends reminders in a polite, escalating sequence, answers routine questions about the invoice, and hands anything contentious to a person. Our research on finance back-office automation found that exceptions and approvals, not data entry, are where finance teams lose the most time, and chasing is the clearest example.
6. Documents that answer with the clause
Advisors spend hours finding the relevant clause in a policy, contract or regulation. An assistant that answers from the firm's own documents and cites the exact clause, so the advisor can check it, turns a search into a question. We covered the contract version in AI contract review. The same pattern works for insurance policies, standard terms and internal procedures.
What is not ready
Three things come up in every conversation and are not where firms should start.
AI giving the advice. A model can draft, summarize and cite. It cannot carry professional liability. The judgment stays with the professional, and the system should make that visible: drafts are labeled, approvals are logged.
Agents sending anything unreviewed. One wrong email to a client costs more than a year of saved minutes. Every agent we build drafts for approval until the firm has seen enough output to decide otherwise. Our guide to human-in-the-loop AI explains the design.
Client data in consumer tools. Confidentiality clauses in your engagement letters apply to what staff paste into a free chatbot. Our guide to keeping business data private with AI has the checklist.
How to pick your first workflow
Score each candidate on three questions. Does it happen many times a week? Is the input already in a system, such as the inbox, the CRM or the document library? Can you measure it with one number, such as response time, proposal hours or days sales outstanding?
The workflow that scores yes on all three is the first project. It usually takes weeks, not months, because it is workflow automation with a model inside rather than a transformation program.
Where this is heading
The next step is one assistant per client relationship rather than one tool per task. It knows the engagement, the documents, the open actions and the history, drafts what each person on the account needs next, and asks before anything leaves the firm.
The pieces exist today in the projects above. What separates the firms that will get there from those that will not is whether their email, CRM, documents and billing are connected, and whether they have agreed who approves what. Both are decisions, not technology.
Frequently asked questions
Will clients object to AI being used on their work?
Most will ask two questions: is our data protected, and does a professional review the output? With a data processing agreement, business-grade AI services and logged approvals, the honest answer to both is yes. Say so in your engagement terms.
Which should we start with, sales or operations?
Whichever has the clearest number. Firms with a long sales cycle usually start with inquiry sorting and follow-ups. Firms with cash-flow pressure start with receivables. Firms that bid for work start with proposals.
Do we need to replace our CRM or practice system?
Rarely. The assistants above read from and write to the systems you have through their APIs. Replacing the system of record is a separate decision and usually a later one.
How long before we see a return?
For a single well-chosen workflow, within a quarter. A follow-up agent or an inquiry sorter shows its effect in the first month's numbers. Firm-wide change takes longer because it depends on connecting systems and agreeing approvals.
What to do this week with AI in professional services
List the five things your fee earners complain about that are not client work. One of them is your first AI project. Our free 5-day AI readiness audit looks at one team and tells you which it should be. See what we build for professional services firms, or book a 30-minute call and talk it through with a founder.
