If you run a 20-person professional services firm, you’ve probably had the AI conversation at least three times this year. Maybe a partner brought it up after reading a trade publication. Maybe a junior associate started using ChatGPT on their own and everybody noticed their output doubled. Maybe a client asked if you’re using AI and you weren’t sure what to say.
The honest answer for most firms this size is: AI can make a meaningful difference, but probably not in the ways you’ve been told. It won’t replace your senior consultants or your experienced CPAs. It won’t write court briefs that stand on their own. But it can take a real bite out of the repetitive work that eats your team’s hours, and at this firm size, every reclaimed hour matters because you’re billing for most of them.
Here’s a realistic walkthrough of where AI fits at a 20-person professional services firm (consulting, legal, accounting, architecture, or similar), what it actually costs, and where the payoff is real versus imagined.
The time audit: where hours actually go
Before anything else, you need an honest picture of how your team spends their time. Not the billable hours. The other hours.
At a 20-person professional services firm, the non-billable administrative work typically breaks down something like this. Document drafting and formatting eats 8-12 hours per week across the team. Research and information gathering takes another 10-15. Client communication (emails, updates, scheduling) is 8-10 hours, and internal knowledge management takes 5-8 more. That last category includes things like finding past proposals, looking up how you handled a similar engagement, or asking colleagues where that template lives. Then there’s the general admin: scheduling, data entry, filing. Another 5-7 hours.
That’s 36-52 hours per week of non-billable or low-value work across a 20-person team. Not all of it is automatable, but a surprising amount is. And at typical professional services billing rates of $150-400 per hour, even recovering 15-20 hours per week has a significant financial impact.
Document drafting: the highest-value use case
This is where most professional services firms should start with AI, because the time savings are immediate and the quality is good enough to be useful right away.
Think about what document drafting looks like at your firm. An associate spends 2-3 hours pulling together a first draft of a proposal, a client memo, a status report, or a scope of work. They start from a previous example (if they can find one), modify the structure, fill in the details, format it, and pass it to a partner for review. The partner redlines it, sends it back, and the associate revises.
With AI, the first-draft phase collapses. The associate feeds the AI a brief with the key details (client name, engagement scope, deliverables, timeline, relevant background) and gets a structured first draft in minutes instead of hours. The draft isn’t perfect. It needs editing, it needs the firm’s voice applied, and it definitely needs a human to verify the specifics. But the starting point is 60-70% of the way there instead of starting from a blank page or an outdated template.
For a 20-person firm, document drafting AI typically saves 5-8 hours per week across the team. That’s the single biggest time win, and it’s also the easiest to implement because it doesn’t require any system integration. Your team can start using AI drafting tools today with just a subscription and some prompt guidance.
The catch: AI-drafted documents pick up a generic quality if you’re not careful. They sound polished but lack the specific phrasing and perspectives that make your firm’s work distinctive. The editing step isn’t optional. Train your team to use AI for the structure and the boilerplate, then add the expertise and personality by hand. The firms that skip the editing step end up with documents that all sound the same, and clients notice.
Research and information gathering
This is the second-biggest time sink at most professional services firms, and AI’s impact here is real but uneven.
For straightforward research tasks (summarizing a company’s recent filings, pulling together background on an industry, reviewing a set of documents for specific clauses), AI is genuinely fast. What used to take a junior associate two hours of reading and note-taking can now be done in 20 minutes with the right AI tool and the right prompts.
The specific use case depends on your firm type. Accounting and tax firms get a lot of value from regulatory research: “Summarize the recent changes to Section 174 R&D expense rules” or “What are the compliance requirements for a company with operations in Texas and California?” produce useful starting points that would have taken significant research time. Consulting firms find it most useful for market research summaries, competitive analysis, and synthesizing large volumes of interview notes or survey responses.
Legal research is where AI is improving fastest, and also where the risks are highest. AI can identify relevant case law, summarize long documents, and flag potential issues in contracts. But AI legal research still requires careful verification. The risk of hallucinated citations is lower than it was two years ago, but it hasn’t been eliminated. Every AI-generated legal reference needs to be checked by someone who knows how to check it.
Across all firm types, expect AI to save 4-7 hours per week on research tasks. The quality gain is less about the AI being smarter than your team and more about it being faster at processing large volumes of text.
Client communication
Professional services is fundamentally a relationship business. Clients choose your firm because they trust specific people. That means you can’t automate the relationship. But you can automate the logistics around it.
AI handles email drafting well. A partner who spends 45 minutes each morning writing status update emails to five clients can feed the AI each project’s current status and get draft updates in 10 minutes. The partner still reviews and personalizes each one, but the time savings add up. Across a 20-person firm, this typically saves 3-5 hours per week.
Meeting scheduling and follow-up is another good fit. AI scheduling assistants can coordinate across multiple calendars, send reminders, and even generate meeting agendas based on recent project activity. After meetings, AI can draft follow-up summaries from notes or transcripts.
Where AI does not help with client communication: anything that requires reading the room, navigating a difficult conversation, or exercising judgment about what to say (and what not to say). A client who’s frustrated about a project delay needs a thoughtful phone call from someone who understands the situation, not an AI-generated apology email. The human judgment piece is exactly what clients are paying professional services rates for.
Internal knowledge management
This one is underappreciated. At a 20-person firm, the institutional knowledge problem is acute. Most of the firm’s expertise lives in people’s heads and in scattered documents across shared drives.
“How did we handle the pricing for that similar engagement last year?” “Where’s the template for the federal compliance report?” “Who on the team has experience with healthcare clients?” These questions get asked constantly, and the answers usually involve interrupting someone who’s trying to do billable work.
An AI-powered internal knowledge base (what we’ve written about in detail elsewhere) changes this dynamic. Your team’s proposals, engagement letters, project reports, and internal memos get indexed and made searchable in plain language. Instead of asking a partner during their lunch break, an associate can ask the system and get an answer with citations to specific documents.
For a 20-person firm, this typically saves 3-5 hours per week in reduced interruptions and faster information retrieval. But the real value isn’t just the time savings. It’s the fact that junior team members can access the firm’s accumulated expertise without always needing a senior person to translate it for them. That changes how fast people ramp up on new engagements.
The honest trade-off: building an internal knowledge base requires organizing your documents first. If your shared drive is a mess (and at most 20-person firms, it is), the setup process includes a cleanup phase that takes 20-40 hours of someone’s time. That’s a real investment, but it pays dividends beyond just the AI system.
Scheduling and general admin
The remaining time savings come from the small stuff that adds up. Meeting scheduling, timesheet reminders, travel coordination, basic data entry.
These tasks individually seem trivial to automate. But across a 20-person team, they consume 5-7 hours per week collectively. AI scheduling tools, email automation, and basic workflow automation can reclaim 2-4 hours of that. Not life-changing per person, but it adds up across the firm.
I’ll be honest: this is the category where the ROI calculation is weakest. The tools work fine, but the time savings per person are small enough (15-20 minutes per day) that people don’t always notice the difference. If you’re choosing where to invest first, start with document drafting and research. Come back to scheduling automation later.
What it costs
Let’s add it up for a 20-person professional services firm.
AI tool subscriptions (writing assistants, research tools, scheduling automation) run $100-300 per person per month, depending on the tools. For 20 people, that’s $2,000-6,000 per month, or $24,000-72,000 per year. Not every person needs every tool. A realistic starting deployment might cover 10-12 power users at $200 per month each, which puts you at $24,000-29,000 per year.
If you’re building an internal knowledge base, add a one-time setup cost of $8,000-15,000 and ongoing maintenance of $200-500 per month.
Training your team on AI tools takes 2-4 days initially, plus ongoing coaching. Budget $3,000-5,000 for an initial training program, then plan for refresher sessions quarterly.
Total first-year investment for a 20-person firm: roughly $35,000-50,000 including tools, setup, and training.
Total time savings: 15-25 hours per week across the firm, with the bulk coming from document drafting and research. At an average billing rate of $200 per hour (conservative for most professional services), that’s $156,000-260,000 per year in recovered billable capacity. Not all of that translates directly to revenue (your team won’t bill every reclaimed hour), but even if you capture 30-40% of it, the investment pays for itself within the first year.
What’s not worth the investment (yet)
Not everything at a professional services firm should be automated. Here’s where I’d tell most 20-person firms to hold off.
AI-generated client deliverables without heavy editing aren’t worth the risk. A consulting report or legal brief that reads like it was written by AI will erode client confidence faster than any efficiency gain can offset. Use AI for drafts. But the final product needs your team’s expertise and voice, always.
Don’t bother trying to automate the complex judgment calls either: which engagement to pursue, how to price a proposal, whether to push back on a client request. You have 20 people. The partners making these calls have deep context that no AI tool can replicate.
Client-facing chatbots are usually a bad fit for professional services, full stop. Your clients are paying premium rates for access to smart humans. A chatbot that answers their questions (even correctly) sends the wrong signal about how you value the relationship. And automated billing systems sound appealing but are harder to implement than vendors suggest, especially with complex engagement structures and custom payment terms. For a firm this size, the setup cost rarely justifies the savings.
The realistic first 90 days
If you’re a 20-person firm considering AI adoption, here’s what a realistic timeline looks like.
Month one: Start with AI writing tools for document drafting. Get 5-8 team members using them for proposals, memos, and client communications. Focus on training and developing prompt templates specific to your firm’s work. Time investment: 2 days of training plus 30 minutes per person per day getting comfortable. Expected savings by end of month one: 3-5 hours per week.
Month two: Add research tools and start building prompt libraries for common research tasks. Begin the document inventory for an internal knowledge base if that’s in scope. Expected savings by end of month two: 8-12 hours per week.
Month three: Launch the knowledge base (if applicable), add scheduling automation, and expand AI writing tools to the full team. Run a retrospective on what’s working and what isn’t. Expected savings by end of month three: 15-25 hours per week.
The firms that do this well don’t treat AI as a technology project. They treat it as a training and workflow project that happens to involve technology. The tools are the easy part. Teaching 20 people to work differently is the hard part. But for firms where time is quite literally money, the math works out clearly enough that it’s hard to argue against trying.