Figma’s stock jumped 37.4% last month, and most of the write-ups pointed at the same thing: enterprise buyers are finally comfortable putting AI agents inside real workflows because the access controls have caught up. That’s the part that matters for a firm like yours, more than the stock price ever will. The technical name for it is “presence,” a way of letting an AI agent operate inside a system while a human decides exactly what it can see and touch. For an accounting or bookkeeping firm sitting on decades of client financial data, that distinction is the whole ballgame.

You’ve probably already had the internal conversation. Someone on your team wants to use AI to draft client emails, summarize a general ledger, or answer a routine question at 9pm on a Sunday. The upside is obvious. The hesitation is also obvious: your firm handles payroll data, bank feeds, tax IDs, and personal financial details for hundreds of clients, and nobody wants an AI tool that can see all of it at once with no fence around it.

Presence-style guardrails solve that specific problem. They let you deploy an agent to do real work while limiting its data access to exactly what that task requires. A close agent sees bank and AP feeds for the entities it’s reconciling. An onboarding agent sees only the new client’s intake documents. Nothing bleeds across engagements, and nothing sits in a shared context window where it could surface in an unrelated conversation. That’s not a compliance nice-to-have anymore. It’s becoming the baseline expectation from clients who ask about your AI policy before they sign an engagement letter.

The manual work this actually targets

Strip away the buzzwords and this comes down to three recurring drains on a firm doing $1M to $25M in revenue.

Month-end and year-end close eats a disproportionate share of your calendar. For most firms we work with, 30% to 50% of total staff time gets concentrated into about four weeks a year. Your best people spend that stretch pulling bank statements, chasing AP discrepancies, and manually keying journal entries instead of doing anything a client would pay a premium for.

Client onboarding is its own tax on growth. Document collection, chart-of-accounts setup, and historical books clean-up routinely stretch into weeks, sometimes longer if the prior bookkeeper left a mess. We typically see 20% to 30% of new clients delay their first billable engagement by a full quarter simply because onboarding drags. Some of them get frustrated enough to walk before you’ve billed a dollar.

And advisory work, the conversations that actually build a firm’s margin, gets crowded out entirely. Compliance work has hard deadlines. Advisory doesn’t. So advisory loses, every time, even though it typically bills at 2 to 3 times the rate of compliance work. Across a firm this size, that combination of month-end drag, onboarding churn, and lost advisory hours usually adds up to somewhere between $60,000 and $180,000 a year in leakage. Not fraud, not waste exactly. Just capacity that quietly disappears.

What a data-guarded agent actually does

Here’s where the presence model changes what’s possible. Instead of one general-purpose AI tool with broad access to everything in your practice management system, you deploy specific agents scoped to specific jobs.

The Month-End Close Agent pulls bank, AP, AR, and payroll feeds for the entities on this month’s close list, reconciles the accounts, flags variances above whatever threshold you set, drafts the journal entries, and assembles a partner-ready close pack. It never touches client files outside the current close cycle. It doesn’t have access to tax return data or engagement letters. Its permissions expire when the close is done and get reissued the next cycle.

The Client Onboarding Agent works the other end of the process. It runs new clients through a guided document collection workflow, builds out the chart of accounts based on their industry, and produces a clean opening trial balance. Its data access is scoped to that one client’s intake folder. It has no visibility into your existing book of business, which matters a lot when you’re onboarding a client who happens to compete with someone you already serve.

The Advisory Insights Agent reads a client’s monthly numbers, surfaces the three things worth discussing in the next meeting, and drafts talking points for the partner walking into that room. This one needs broader visibility into a single client’s trend data, but it’s still fenced to that client. It doesn’t cross-reference against other accounts, and it doesn’t retain anything once the meeting prep is delivered.

That’s the actual shape of “presence” applied to your firm. Not one AI that knows everything. A set of agents, each with a job, each with a boundary, and a partner who can see exactly what each one accessed and when. If you want a longer walk-through of how these three agents fit together across a full month-end cycle, our Month-End AI Close Map for Accounting Firms lays out the workflow step by step, including where the handoffs between agent and staff happen. It’s built as a practical worksheet, not a sales deck, and it’s meant to be used with your existing close checklist rather than replacing it.

Why this matters more for accounting firms than most industries

Retailers and marketing agencies get to experiment with AI access controls at a leisurely pace. You don’t have that luxury, because the downside of a data leak in this industry isn’t embarrassing. It’s a professional liability event. A single instance of an AI tool surfacing one client’s financial data inside a conversation about another client could end an engagement, trigger a disclosure obligation, or worse.

That’s exactly why the presence model is worth taking seriously rather than dismissing AI adoption altogether. The old choice was binary: keep everything manual and safe, or open things up and hope nothing goes wrong. Granular access control breaks that binary. You can automate the month-end grind, speed up onboarding, and free up advisory time, without giving any single tool a master key to your client data.

We built our Omni ops agents around this exact principle, and it shows up in how we scope every deployment for accounting and bookkeeping clients. Each agent gets read access to the specific feeds, folders, or ledgers its task requires. Nothing more. You approve the scope before it goes live, and you can audit what it touched afterward. It’s closer to how you’d manage a new junior staff member’s access on day one than how most people picture “AI.”

What this looks like in your calendar, not just your ledger

Put a number on it. If your firm has 12 people and you’re losing 35% of staff capacity to month-end crunch weeks, that’s roughly four to five weeks a year where almost nobody is doing anything except close work. At a blended billable rate, that’s real money sitting idle even when the work gets done, because it’s not advisory-rate work.

Now add onboarding drag. If a quarter of your new clients don’t generate billable hours for three months, and your average new client is worth $15,000 to $40,000 annually depending on service tier, you’re deferring a meaningful chunk of revenue every single year, plus absorbing the churn risk of clients who get frustrated and leave before they ever became profitable.

Then there’s the advisory gap. If compliance work fills every available hour and advisory conversations happen for maybe 10% of your client base instead of 50%, you’re leaving the highest-margin part of your practice almost entirely on the table. Multiply a modest per-client advisory fee across the clients who never get the conversation, and it stacks up fast.

None of these numbers are dramatic on their own. Together, for a firm in the $1M to $25M range, they tend to land in the $60,000 to $180,000 annual range we see across this vertical. That’s not a rounding error. That’s often the difference between a firm that can hire its next senior associate and one that can’t.

Where to start without betting the firm on it

You don’t need to deploy three agents across your entire practice on day one. The firms that get this right start narrow. Pick the close, or pick onboarding, run it for one engagement cycle, and watch what the agent actually touches versus what you expected it to touch. Adjust the scope. Then expand.

That’s also roughly the shape of what we do in an Omni Audit. It’s 60 minutes, no slide deck, and you walk away with three concrete things: a map of where your firm’s hours are actually going right now, a scoped recommendation for which agent to deploy first, and a rough dollar estimate of what’s leaking under your current process. If you want to see how this plays out for firms your size, see Omni for accounting and bookkeeping before you commit to anything. It’s a look at the numbers, not a pitch.

We also cover more of the mechanics of how firms are structuring advisory work differently once compliance stops eating the whole calendar in our advisory resources, and if you’re curious how other service businesses are handling the same access-control questions, our insights section has a handful of adjacent write-ups. For a broader sense of how the underlying agent platform works before you talk to anyone, /omni is the place to start.

The actual next step

If you’re running a firm in this size range and month-end still feels like a fire drill four times a year, or you’ve watched a new client’s onboarding stretch from three weeks into three months, the fix isn’t a bigger AI subscription. It’s a scoped deployment that respects the boundaries your clients expect you to respect anyway.

Book my Omni Audit and we’ll walk through exactly where your firm stands, using your own numbers rather than industry averages. Sixty minutes, three concrete outputs, and you’ll know within the hour whether the Month-End Close Agent, the Client Onboarding Agent, or something else entirely is the right first move.

Presence controls made headlines this month because a design software company proved enterprises will pay for AI they can actually govern. Your clients are watching the same trend, whether they say so out loud or not. Firms that can show a client exactly what data an agent touched, and what it didn’t, are going to win the engagements that firms with a vague “we use AI” answer will lose. If you want a second opinion on where your firm sits on that spectrum, the AI audit for accounting and bookkeeping is still the fastest way to find out, and our guides section has more detail on how the scoping process works once you’re ready to move past the audit stage.