Microsoft recently disclosed that custom AI agents built on its platform now outnumber Copilot licenses, with more than 40 million agents deployed across its customer base. That’s not a small signal. It means the market has moved past “give everyone a chat assistant” and into “build agents that do one job extremely well, repeatedly, without a person driving them.”

For consulting and advisory firms, this shift matters more than it does for most industries. Your business is built on repeatable intellectual work that somehow never gets treated as repeatable. Every proposal starts from a blank deck. Every engagement kicks off with research someone already did for a different client eighteen months ago. Every project produces documents that get buried the moment the invoice clears. A general-purpose assistant like Copilot can help someone write faster. It can’t fix any of that, because it doesn’t know your firm’s history, your pricing logic, or what’s sitting in last year’s project folders.

That’s the gap custom agents close. Not a chatbot bolted onto Office. A purpose-built worker trained on your firm’s own material, running the same workflow every time, without needing a senior person to babysit it.

Why Copilot alone doesn’t solve this

Copilot and tools like it are genuinely useful for drafting an email or summarizing a document you hand it directly. But they start from zero every time. They don’t have a memory of your last 50 proposals. They can’t tell you which pricing model won you the last three engagements in manufacturing. They won’t cross-reference a client’s industry against the eleven research briefs your team already produced this year.

That’s the distinction Microsoft’s own numbers are pointing at. Firms aren’t just buying more seats of a general tool. They’re building narrow, task-specific agents that plug into their own data and repeat a defined process. For a consulting firm, the defined processes are obvious once you list them out. Proposal writing. Client and industry research. Turning finished project work into something the next team can actually use.

We built our Omni ops agents around exactly that logic, and it maps almost perfectly onto where consulting firms bleed the most time and money.

The three places the work repeats and nobody notices

Proposals eat senior time that should be billable

Ask any partner how long a major proposal takes and you’ll hear something in the 20 to 40 hour range, spread across the team but concentrated in the people who are also supposed to be running client work. Someone has to pull the right case studies, rewrite the approach section for the new client’s language, get pricing approved, and format the whole thing so it looks sharp. None of that is thinking work. It’s assembly work dressed up as thinking work, and it happens every single time a new opportunity lands.

Firms with a healthy pipeline end up penalized for it. The busier you are, the more proposals you’re writing, and the more senior hours get pulled away from delivery and into deck production. Win rates might be fine. The cost of winning is what quietly erodes margin.

Research gets repeated across clients without anyone tracking it

Most engagements open with a research phase. Someone spends one to three weeks building an industry view, mapping competitors, or synthesizing regulatory changes before the actual advisory work starts. That’s normal and reasonable the first time. The problem is that a consulting firm serving multiple clients in adjacent sectors ends up doing variations of the same research over and over, and almost none of it gets reused, because it’s saved in a folder tied to one client engagement and nobody thinks to check it before starting the next one.

That’s not thin waste. Across a firm doing $1M to $25M in revenue, we typically see this kind of duplicated research compounding into a meaningful chunk of the $80,000 to $300,000 a year in avoidable cost that shows up when you actually audit where hours go.

Every project makes IP that the firm never gets to use again

This is the one owners feel but rarely quantify. Every engagement produces decks, working documents, interview notes, and a final report full of frameworks and insight that took real expertise to build. Then the project closes, the team moves on, and that material sits in SharePoint or Google Drive, technically searchable but practically invisible. Six months later a different team hits a similar client problem and starts from scratch, because nobody remembers the first project existed, let alone what it concluded.

The firm ends up paying for the same insight twice. Once to produce it, and again to reproduce it because it wasn’t captured anywhere a person would actually think to look.

What a purpose-built agent actually does, end to end

This is the part that gets lost in the noise around “AI agents” as a buzzword. A well-built agent isn’t a smarter chatbot. It’s a defined process running on your firm’s own material, with clear inputs and a clear output. Here’s what that looks like for the three problems above.

The Proposal Generation Agent pulls from your firm’s actual proposal history, case studies, and pricing structures. Feed it the new opportunity, the client’s industry, and the scope discussed on the call. It drafts a tailored proposal using the language, structure, and pricing logic that’s already worked for you, pulling the right case studies automatically instead of someone digging through old folders trying to remember which client story fits best. Your team edits and refines. They don’t start from a blank page.

The Research Agent runs at the start of every engagement. Give it the client name and industry and it produces structured research with sources, summaries, and a one-page brief the team can actually use in the first client meeting. It also checks against research the firm has already produced, so nobody spends three weeks rebuilding an industry view that already exists in a report from last year.

The Knowledge Agent reads every deck, document, and meeting transcript the firm generates, and lets anyone on the team ask a question across that entire corpus. Instead of pinging six people to ask “did we ever do work in this vertical,” someone just asks the agent, and it points to the specific project, the relevant framework, and the person who led it.

None of these are meant to replace judgment. Every output still gets reviewed by someone senior before it goes to a client. What they replace is the hours of assembly, retrieval, and re-creation that happen before judgment is even applied. That’s the part Copilot, used on its own, was never designed to fix, because it doesn’t sit inside your workflow with access to your firm’s own history.

What this is worth to a firm your size

If your firm does $1M to $25M in revenue, the leakage from these three problems typically lands in the $80,000 to $300,000 a year range once you count senior hours spent on assembly work, duplicated research across engagements, and the rework that happens when knowledge doesn’t transfer between projects. That’s not a hypothetical. It’s what shows up when you actually map hours against outcomes, which most firms never do because everyone’s too busy delivering client work to audit their own.

The math is straightforward even without precision. A partner billing at $300 to $500 an hour spending 30 hours on a proposal is a $9,000 to $15,000 cost of sale before the client ever says yes. Do that eight or ten times a year and you’re looking at a six-figure number attached to work that doesn’t require partner-level thinking, just partner-level access to the firm’s history.

If you want a broader view of how firms in professional services are approaching this shift, our resources on AI strategy cover it from a few different angles, and the guides section has more detail on how agent-based workflows differ from general AI tools in practice.

Where to start without overbuilding

You don’t need to automate everything at once, and you shouldn’t try. The firms that get real value out of this start with the single workflow that costs them the most senior time, prove it works on one or two engagements, then expand. Proposal generation is usually the highest-leverage starting point because the time cost is visible and the win is fast. Research and knowledge management tend to follow once the first agent earns trust.

If you want a structured way to think through that first build, we put together a practical worksheet called Deploy Your First Business Agent, which walks through how to pick the right first workflow, what data it needs access to, and how to measure whether it’s actually saving time. You can grab the direct download here if you’d rather skip the landing page.

That said, a worksheet only gets you so far. The firms that move fastest are the ones that get a clear, honest look at where their specific hours are going before they build anything.

The Omni Audit, and why it beats guessing

We run something we call the Omni Audit specifically for firms in this position. It’s 60 minutes, on a call, no deck. We walk through your actual proposal process, your research workflow, and how project knowledge moves, or doesn’t move, across your team. You walk away with three concrete outputs, a map of where your hours are actually leaking, a rough dollar estimate of what that leakage costs annually, and a prioritized list of which agent to build first based on your specific bottleneck, not a generic template.

No slide deck, no sales pitch dressed up as strategy. Just an honest look at whether your firm’s version of the proposal problem, the research problem, or the knowledge problem is costing you the most, and what building against it would actually look like. You can see Omni for consulting firms to get a sense of how we approach this before you book anything.

If you’re already fairly confident where your firm’s version of this problem sits, the fastest next step is to book a 60-min Omni Audit and we’ll map it against your numbers directly.

The broader signal is worth taking seriously

Microsoft’s own data is a useful checkpoint here, not because the exact figure matters to your firm, but because of what it tells you about where the market is heading. Enterprises with far more resources than a $1M to $25M consulting firm are moving past general-purpose AI tools and into narrow, purpose-built agents tied to their own workflows and data. That’s not a trend consulting firms can sit out, given how much of your cost structure is tied up in repeatable intellectual work that currently gets redone by hand every time.

The firms that build their own proposal, research, and knowledge agents now will be running leaner engagements a year from now, with senior time going toward client thinking instead of deck assembly. The ones that wait will keep paying the $80,000 to $300,000 a year quietly, spread across proposals nobody tracks the hours on, research nobody reuses, and IP nobody can find.

If you want to know which bucket your firm falls into, that’s exactly what the audit is for. You can explore how Omni’s ops agents work in more depth, or go straight to seeing Omni for consulting firms and book a slot when you’re ready. Either way, the 40 million agent number from Microsoft isn’t really about Microsoft. It’s a signal that the firms winning the next few years are the ones that stopped treating repeatable work as unavoidable overhead.

When you’re ready to put real numbers against your own firm, book a 60-min Omni Audit and we’ll spend the hour on your actual workflows, not a generic pitch.