A firm we talked to recently ran an internal audit before budget season, expecting to find maybe four or five AI subscriptions floating around the practice. They found closer to forty. Different partners had signed up for different transcription tools. A senior bookkeeper was paying for a personal ChatGPT Plus account and expensing it. Two teams were running the same document-extraction tool under different vendor names. Nobody had planned any of this. It just happened, one credit card swipe at a time.

This is Phase 1 of what’s becoming a familiar pattern: buy all the AI. Nobody plans Phase 2. Phase 3, the profit part, never shows up because nobody consolidated, nobody measured usage, and nobody connected the tools to actual firm workflows like month-end close or client onboarding.

If you run an accounting or bookkeeping firm doing $1M to $25M in revenue, there’s a good chance a version of this is already happening inside your practice. Not because your team is careless, but because AI tools are cheap enough and easy enough to sign up for that individual staff members solve their own pain points without telling anyone. The result is a shadow AI stack that costs real money and produces almost none of the leverage a coordinated system would.

What a shadow AI audit actually finds

Start with the obvious question: how many AI tools does your firm currently pay for? Most partners we talk to guess low. They think about the one or two tools leadership approved, maybe a transcription app for client calls or a categorization assistant bolted onto the accounting software. What they miss is everything staff bought on their own.

A typical shadow audit for a firm this size turns up a few consistent categories:

• Personal subscriptions expensed as “software” with no owner or renewal tracking
• Duplicate tools solving the same problem for different teams, often at different price points
• Free-tier tools that quietly upgraded to paid tiers nobody reviewed
• Point solutions bought during tax season crunch that never got cancelled after April

None of this shows up as one big line item. It shows up as $40 here, $89 there, spread across a dozen vendors and a dozen credit cards. Add it up across a firm with 15 to 40 staff and you’re often looking at $60,000 to $180,000 a year in unmanaged AI and software spend, much of it doing overlapping work that a coordinated system could handle for a fraction of the cost.

The deeper problem isn’t the subscription fees themselves. It’s that none of these tools talk to each other, none of them are built around your actual bottleneck workflows, and none of them free up the hours that matter most: the four weeks around month-end and year-end when 30-50% of staff time gets consumed by a crunch everyone saw coming.

Where the manual work actually lives

Shadow AI tends to cluster around three points in a firm’s calendar, and it’s worth naming them plainly because they’re the same three points where a properly built agent earns its keep.

The first is the month-end and year-end close. This is the predictable spike nobody plans around well. Staff are pulling bank feeds, matching AP and AR, chasing payroll numbers, and building journal entries under deadline pressure. Somebody on your team probably bought an AI tool specifically to speed up part of this, maybe bank reconciliation or receipt categorization, and it helps a little. But it’s a patch, not a system. It doesn’t talk to your general ledger workflow, doesn’t flag variances against prior periods, and doesn’t produce anything a partner can actually review before the client call.

The second is client onboarding. Document collection, chart-of-accounts setup, and historical clean-up routinely take weeks longer than anyone quotes the client. Firms in our network tell us 20-30% of new clients delay their first billable engagement by a quarter simply because onboarding drags. Somewhere in that mess, someone probably bought a document-collection tool to speed things up. It helps with one slice of the problem and does nothing for the chart-of-accounts setup or the opening trial balance, so the drag continues.

The third is advisory time. Compliance work eats the calendar because it has hard deadlines and clients notice when it’s late. Advisory conversations, the ones billing at 2-3x the compliance rate, get pushed to “when things calm down,” which for most firms is never. This is the part of the practice shadow AI tools almost never touch, because nobody bought a tool with advisory in mind. They bought tools to survive the crunch, not to create space for the higher-margin work.

If you want a structured way to walk through where your own close process leaks time, the Month-End AI Close Map for Accounting Firms is a practical worksheet built for exactly this. It maps the close process week by week and flags where duplicated tools or missing automation are costing you hours. You can download it directly here and use it before your next planning conversation.

What a real agent looks like end-to-end

The fix isn’t cancelling every subscription and starting from zero. It’s consolidating around a small number of agents built for your actual workflows, so the tools staff were improvising with get replaced by something that does the whole job, not a slice of it.

Take the Month-End Close Agent. Instead of a patchwork of reconciliation apps and manual spreadsheet checks, this agent pulls bank, AP, AR, and payroll feeds directly, reconciles them, flags variances against expected ranges, drafts the journal entries, and prepares a close pack that’s ready for partner review. Staff aren’t spending the last week of the month re-keying numbers or hunting for the discrepancy in a client’s payroll run. They’re reviewing an output that’s already 90% done and spending their time on the judgment calls that actually need a human.

The Client Onboarding Agent works the same way on the front end of the client relationship. It runs a guided workflow that collects documents from new clients, sets up the chart of accounts based on the client’s industry and structure, and produces a clean opening trial balance without three weeks of email back-and-forth. Firms using this kind of agent tend to see onboarding timelines shrink from weeks to days, and the churn that happens when a new client gets frustrated waiting for their first deliverable drops with it.

The Advisory Insights Agent is the one that directly attacks the third pain point. It reads each client’s monthly numbers, surfaces three specific things worth discussing, and drafts the partner’s talking points before the meeting happens. This doesn’t replace the advisory conversation. It removes the two hours of prep that usually stops the conversation from happening at all. Partners walk into the meeting with something specific to say instead of a generic “how’s business” opener, and clients start to see the advisory relationship as worth paying for.

None of these three agents require staff to sign up for their own tools and expense them quietly. They’re built once, they connect to the systems you already run, and they replace the shadow stack with something the whole firm can see and manage.

The dollar math behind the audit

It’s worth being specific about what this actually costs a firm your size, because “AI waste” sounds abstract until you put a number next to it.

Industry ranges we typically see for firms with $1M to $25M in revenue put unmanaged AI and adjacent software spend somewhere between $60,000 and $180,000 a year. That figure includes duplicate subscriptions, unused seats, tools bought for a single tax season and never cancelled, and the labor cost of staff manually stitching together tools that don’t talk to each other. It doesn’t even count the opportunity cost of the advisory hours that never happen because compliance work fills the calendar.

That gap between what leadership thinks is being spent and what’s actually being spent is the whole reason Phase 2 never happens. You can’t consolidate what you can’t see. And you can’t build the case for investing in something like the Month-End Close Agent or the Advisory Insights Agent if the current spend is invisible and scattered across forty different vendor relationships.

This is exactly the kind of thing worth reviewing before tax season spending ramps up again, not after. If you’re serious about entering the next busy season with a consolidated stack instead of another round of one-off tool purchases, now’s the time to get the audit done.

What an Omni Audit actually covers

We built the Omni Audit specifically for firms in this spot: you suspect there’s waste, you know your close and onboarding processes have friction, but you don’t have a clean picture of where the dollars and hours are actually going.

It’s 60 minutes, and it produces three concrete outputs. First, a map of your current AI and software spend, including the subscriptions nobody remembers signing up for. Second, a breakdown of where your close, onboarding, and advisory workflows are losing time, tied to actual hours and actual dollar ranges for a firm your size. Third, a prioritized list of which agents would generate the fastest return if you built them first, whether that’s the Month-End Close Agent, the Client Onboarding Agent, or something specific to your practice.

No deck. No generic AI strategy presentation. Just a working session that gives you the same visibility that firm we mentioned earlier got when they finally counted their subscriptions and found ten times more than they expected.

If you want to see how this applies specifically to your practice, see Omni for accounting and bookkeeping before you book, or go straight to the AI audit for accounting and bookkeeping to get a sense of what the audit covers for firms your size. Either way, the next step is the same conversation.

Before you buy the next tool

The instinct when a new AI tool promises to save time during close or speed up onboarding is to just buy it and see. That instinct is exactly how firms end up with forty subscriptions and no coordinated system. Before the next purchase, it’s worth asking whether the problem is a missing tool or a missing workflow. Most of the time, once you’ve done a proper audit, it turns out to be the second thing.

If you want to read more about how firms in adjacent industries are handling this same shadow AI problem, our insights section covers a few different angles, and our broader guides library has more detail on how the Omni ops agents get built and deployed inside a firm like yours. You can also look at Omni ops directly to see how the close, onboarding, and advisory agents fit together as one system instead of three separate purchases.

The firms that get ahead of this aren’t the ones buying the most AI tools. They’re the ones who audited what they already had, cut the duplicates, and built two or three agents that actually touch their biggest bottlenecks. That’s a very different outcome than Phase 1 buy-everything, and it’s the one that shows up in your margin instead of your subscription list.

If you’re ready to find out what your own shadow AI stack is costing you, book a 60-min Omni Audit and we’ll walk through your actual numbers together. No generic pitch, just your spend, your workflows, and a clear next step. You can also book my Omni Audit directly if you already know this is worth 60 minutes of your time before tax season spending starts again.