Every partner meeting I sit in on eventually gets to the same question. Should we build our own AI tools, or just buy something off the shelf? Usually someone has a nephew who “does AI” or a junior consultant who’s been playing with agent frameworks on the side, and the room splits into two camps. One wants to commission a custom build. The other wants to wait and buy whatever the big vendors ship next quarter.

Both camps are half right, and that’s the problem. The real answer isn’t build or buy. It’s build and buy, applied to different parts of the business with a clear line between them.

The build vs buy question every consulting firm gets wrong

Here’s the mistake I see most often in firms doing $1M to $25M in revenue. They either try to build everything in-house because they don’t trust vendors with client data, or they buy a generic AI platform and expect it to understand their methodology out of the box. Neither works.

The framework that actually holds up is simple. Buy AI for anything that looks the same at every consulting firm on earth. Build AI for the thing that makes your firm different from the consulting firm down the street.

Scheduling, CRM hygiene, expense processing, basic email triage, meeting transcription. These are commodity workflows. Twenty vendors have already solved them better than your firm will solve them internally, and cheaper. There’s no competitive advantage in having a slightly better internal calendar bot. Buy that.

But the way your firm structures a proposal, the frameworks you’ve built over a decade of engagements, the research shortcuts your best consultants have in their heads and nowhere else, that’s proprietary. That’s the stuff a generic tool can’t replicate because it doesn’t know your firm’s IP. That’s where custom agent work earns its cost.

What’s actually eating your senior people’s time

Before you decide what to build, it helps to know where the hours actually go. In firms this size, three things show up over and over.

The first is proposal and pitch time. A senior partner sits down to write a major proposal from scratch, pulls old decks from a shared drive that may or may not be organized, guesses at pricing based on memory, and burns 20 to 40 hours getting a document out the door. Win rate is usually fine. The cost of getting to that win is the problem. That’s partner-level time spent formatting slides instead of talking to prospects.

The second is research and synthesis. Every new engagement starts with two or three weeks of secondary research, industry scans, competitor reviews, market sizing. It’s necessary work, but a huge portion of it repeats across clients in the same sector. Nobody keeps it in a form the next team can reuse, so the firm pays for the same research effort again and again, project after project.

The third is knowledge management debt, and it’s the quiet one. Every engagement generates real intellectual property, frameworks, findings, client-specific adaptations of your methodology. Almost none of it makes it back into a form anyone else in the firm can find or use. Ask a partner where last year’s retail-sector findings live and you’ll get a shrug or a folder name nobody remembers. The firm effectively pays for the same insight twice, once to generate it and again when someone rebuilds it from scratch six months later.

None of these three problems get solved by buying a better calendar tool. And none of them get solved by hand-coding a custom scheduling agent nobody asked for. They sit squarely in the “build” column, because they touch the actual IP of the firm.

The case for buying the standard stuff

I want to be direct about this because it’s where firms waste the most money on the “build” side. If your ops team wants to build a custom Slack bot for meeting reminders, or a bespoke CRM sync tool, stop them. Buy it.

The math is straightforward. A commodity AI tool for scheduling or CRM enrichment costs a few hundred dollars a month and works on day one. A custom build for the same thing costs tens of thousands in developer time, takes months, and needs maintenance every time the underlying model updates. You’re paying a premium to reinvent something a vendor already sells well.

This is the single biggest source of wasted spend I see when firms first start exploring AI. They hire a developer, or worse, a whole agency, to build something generic that a $50-a-month tool already does. If you want a gut check on which of your current workflows fall into this bucket versus the ones worth custom-building, that’s exactly the kind of triage we walk through in the AI audit for consulting firms. It usually takes less time than the meeting where you’re currently debating it.

Where building actually makes sense

Once the commodity stuff is off the table, what’s left is usually three types of work, and they map almost exactly to the three pains above.

Proposal Generation Agent. This is a build, not a buy, because your proposals aren’t generic. The agent needs to pull from your firm’s actual past proposals, your actual case studies, your actual pricing logic, and produce a tailored first draft for the specific opportunity in front of you. A generic AI writing tool can produce a proposal-shaped document. It can’t produce a proposal that sounds like your firm and prices like your firm. That gap is exactly where custom work pays for itself. Firms running this well cut proposal drafting from the 20 to 40 hour range down to a few hours of partner review and editing, which frees senior time for the parts of the sale that actually need a senior person.

Research Agent. At the start of every engagement, this agent runs structured industry and company research, pulls sources, produces summaries, and delivers a one-page brief before the kickoff call. It’s built specifically to know your firm’s research standards, your preferred sources, and the format your consultants actually use, not a generic search summary. Over a year of engagements, the research this agent produces starts compounding because it’s stored and searchable, instead of vanishing into someone’s laptop after the project closes.

Knowledge Agent. This is the one most firms underestimate until they see it work. It reads every deck, every document, every meeting transcript the firm has produced, and answers questions across that whole body of work. A new consultant staffed onto a manufacturing client can ask “what have we found in manufacturing engagements about inventory forecasting” and get an answer pulled from actual past work, with sources, in seconds. That’s the knowledge management debt problem solved directly, not with a better filing system, but with an agent that actually reads and understands what’s already in the firm.

Notice what these three have in common. They’re all trained on your firm’s own material. A vendor selling a generic “AI for consulting” product has never seen your proposals, your pricing model, or your past project files. It can’t. That’s precisely why these three sit in the build column and scheduling doesn’t.

What this looks like end to end

Picture a real engagement kickoff. A partner gets a call from a mid-sized manufacturer looking for a supply chain review. Old workflow: partner assigns a consultant to spend a week and a half pulling industry reports, checking competitor filings, and building a market backgrounder before anyone even talks pricing. New workflow with a Research Agent in place: the partner enters the client name and sector, and within a couple of hours has a sourced one-page brief covering market conditions, likely competitors, and relevant regulatory context. The consultant’s first week is now spent on client conversations, not desk research.

Same engagement, three weeks later. The team’s found something interesting about the client’s warehouse allocation logic. Under the old model, that insight lives in a slide deck on someone’s laptop and probably never gets looked at again. With a Knowledge Agent in place, that finding gets indexed automatically, and six months later when a different partner is scoping a similar client, they can ask the system directly instead of emailing around the firm hoping someone remembers.

And when it’s time to write the proposal for the next phase of work, the Proposal Generation Agent pulls the pricing structure the firm used on the last three similar engagements, drafts the executive summary in the firm’s actual voice because it’s trained on the firm’s actual past proposals, and hands the partner a document that needs editing, not writing from a blank page.

None of this requires replacing your CRM or your scheduling tool. Those stay bought, not built. What changes is the work that touches your actual methodology, and that’s the whole point of drawing the line where we’ve drawn it.

The dollar math for a $1M-25M firm

Run the numbers on your own firm for a second. If your senior people spend even 20 to 40 hours per major proposal, and you’re running a dozen or more significant proposals a year, that’s several hundred hours of partner-level time going into document production instead of client work or business development. At typical partner billing rates, that alone often lands somewhere in the $80,000 to $300,000 range annually once you count the opportunity cost, not just the hours.

Add in the research that gets rebuilt project after project because nothing’s reusable, and the IP that gets generated and then buried, and you start to see why firms this size consistently sit in that same leakage band. It’s not one big obvious cost. It’s three ordinary-looking inefficiencies compounding quietly across every engagement the firm runs.

That number is worth sitting with before you decide anything about build or buy. It tells you how much room there is to justify the investment in the agents that actually touch your IP, and how little sense it makes to spend that same budget building a custom version of something you could buy for a few hundred dollars a month.

Start with an audit, not a build decision

The mistake most firms make is jumping straight to “let’s build an agent” before they’ve mapped where the hours are actually going. You don’t need a roadmap deck or a six-week discovery phase to figure that out. You need someone to sit with your actual workflow for an hour and tell you, honestly, what’s worth buying, what’s worth building, and what’s not worth touching yet.

That’s what an Omni Audit is. Sixty minutes, no slide deck, three concrete outputs, a leakage estimate specific to your firm, and a short list of which workflows are commodity buys versus proprietary builds. If you want a lighter starting point before that call, the Deploy Your First Business Agent worksheet walks through the same build-versus-buy triage on your own time, and it’s a useful way to come into the audit with your own list already sketched out. You can grab the direct checklist here if you’d rather work through it before booking anything.

If you’d like more background before that conversation, our insights section and guides cover how firms in professional services are sequencing these builds, and the Omni Ops page breaks down how the Research, Proposal, and Knowledge agents actually connect to your existing systems rather than replacing them wholesale.

Your next 60 minutes

You don’t have to decide today whether to build or buy anything. What you should decide is whether you actually know, right now, where your firm’s hours and dollars are leaking. Most partners we talk to have a rough sense but not a real number, and that rough sense is usually wrong in one direction or the other.

The audit fixes that in an hour, with no obligation and no deck to sit through afterward. Book a 60-min Omni Audit and we’ll walk through your actual proposal process, your research workflow, and your knowledge management gaps together, then tell you plainly what’s worth buying off the shelf and what’s worth building around your firm’s own methodology.

If you want to see how this plays out specifically for firms your size, see Omni for consulting firms has more detail on what the audit covers and what firms typically find. Either way, the sooner you know your real numbers, the sooner the build-or-buy question stops being a debate and starts being a plan. Book my Omni Audit and let’s get you that plan.