Someone finally built this for the person running the business alone

Anthropic has released an AI package built specifically for small businesses: fifteen ready-built workflows and fifteen skills that plug directly into the software a one or two person operation already runs, from QuickBooks to HubSpot to Google Workspace. It is the first serious attempt this year to build for the owner who is also the bookkeeper, the marketer and the person closing up on a Friday night, rather than for a company with an IT department to lean on.

Most of what gets built in AI is built for someone who already has an IT department to lean on. A workflow that assumes there's someone to set up the integrations. A rollout plan that assumes there's a team big enough to have a training day. A feature list that assumes the person reading it isn't also the one answering the phone, doing the books, and closing up on a Friday night. None of that is malicious. It's just who most software gets designed for by default, and it means the businesses that could arguably benefit most from getting this right, the ones running on one or two people's effort with no spare capacity anywhere, have spent this whole year being handed tools built for someone else's business.

In May this year, Anthropic, the company whose tool sits behind Cowork, said the quiet part out loud about exactly this problem. Announcing a new package built specifically for smaller operations, they put the failure plainly: "Tools and training are rarely tailored to the ways small businesses operate, and as a result their use often stops at the chat window." That's the entire story of AI adoption for most of the businesses I work with, told in one honest sentence by the company that makes the tool. A chat window is a properly useful thing on its own. It is also, for a business this size, roughly a tenth of what's actually available, and most owners have never been shown the other nine tenths because nothing was ever built with them specifically in mind.

Why does AI use stop at the chat window for most small businesses?

Because a chat window is where AI use naturally lands for a business owner with no spare hour to spend working out how to connect a new tool to the accounting software, the CRM, the design platform and everything else the business already runs on. It is not laziness or a lack of curiosity; it is the only kind of AI use that fits in the gaps between everything else.

It's worth being precise about what "stops at the chat window" actually means in practice, because it's easy to hear that phrase and assume it's describing laziness or a lack of curiosity. It isn't. Learning to have a good conversation with an AI tool is useful and entirely achievable in the gaps between everything else. Learning to wire that same tool into six different pieces of business software, safely, with the right permissions in the right places, is a different kind of project entirely, and it's the kind of project that gets pushed to next quarter indefinitely in a business where next quarter never actually arrives with spare time in it.

Anthropic backed this up with a number worth sitting with, even though it's a United States figure rather than an Australian one and the comparison doesn't translate directly: small businesses account for 44 percent of US GDP and employ nearly half the private-sector workforce there, and their AI adoption has still lagged well behind larger enterprises. Take that as a description of a structural pattern rather than a number to import wholesale into an Australian context. The pattern is the point. It isn't that small business owners are less capable or less interested. It's that almost everything built in this category assumes a team with the time and the technical depth to do the wiring themselves, and the businesses without that team get left exactly where the chat window leaves them: using AI for one conversation at a time, while the actual operational leverage sits unused a few steps further on.

What is actually in Anthropic's small business package?

Fifteen ready-built workflows spanning finance, operations, sales, marketing, HR and customer service, plus fifteen additional skills aimed at the specific bottlenecks that show up again and again in businesses this size. None of it requires learning a new platform from scratch; it plugs directly into the software a small business is already running.

The package itself is worth describing plainly, because the detail matters more than the announcement. It plugs directly into the software a small business is already running: Intuit QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, Microsoft 365. The point of naming all of those specifically is that this isn't a generic productivity suite asking a business to migrate its systems to fit the tool. It's built to sit inside the systems already in place, which is precisely the barrier that's been keeping most small businesses stuck at the chat window in the first place.

That distinction matters more than it might sound like it does on first read. A business that has to migrate its accounting or its client records to adopt a new AI tool has, in practice, ruled the tool out before it's even tried it, because the migration itself is the very kind of project that never survives contact with a business that has no spare capacity. A tool that instead plugs into QuickBooks or HubSpot as they already exist removes exactly that barrier. Nothing has to change about how the business already runs before the tool can start being useful inside it.

It's also worth naming what fifteen workflows plus fifteen skills actually spares a business from having to build itself, because the number on its own doesn't convey much. Building even one properly working, properly connected automation from scratch, the kind that pulls real data from a real system and does something useful with it reliably, is not a quick job even for someone who knows what they're doing. It typically means mapping the process precisely enough to hand to a system, choosing the right connection points, testing it against enough real cases to trust it, and then maintaining it as the tool underneath it keeps changing. Thirty pre-built starting points across the six functions that make up most of what a small operation actually does removes the building step entirely for anyone who doesn't need something bespoke. What's left is the far smaller job of checking whether a given workflow fits how this particular business actually runs, and adjusting it, rather than starting from a blank page every single time.

What are the honest limits of a pre-built AI package?

Fifteen workflows built to fit small businesses in general will fit some businesses closely and others only loosely, and the honest expectation is a genuine adjustment period rather than an instant, perfect match. That is not a flaw in the product; it is the ordinary cost of anything built to serve a category rather than one specific business.

None of this is a case for treating a pre-built package as a finished solution the moment it's switched on. A workflow built for a typical services business's invoicing pattern may need real tweaking before it matches how a specific trades business or a specific retail operation actually structures its billing. It's worth saying plainly rather than implying the moment you flip the switch, everything runs exactly as your business needs it to on the first try.

What is the trust dial, and why does it matter?

Anthropic describes the control built into every one of these workflows in a single sentence: "You approve the plan first or, when you're ready, let it run end-to-end." That is, in plain terms, exactly the mid-task checkpoint every business needs, provided as a toggle built into the product rather than something a business owner has to design from scratch.

I want to be careful about what this does and doesn't solve, because it would be easy to read this as the governance question resolved and move on, and that would be the wrong lesson to take from it. Having the toggle built in means a business doesn't have to invent the mechanism for pausing before a task runs unsupervised. It still has to decide, task by task, which setting is the right one, and that decision is exactly as much the owner's job as it was before the toggle existed. What the product removes is the technical excuse. A business can no longer say it doesn't have oversight because building oversight was too hard or too technical to set up. The switch is right there. What's still required, and always was, is a person willing to actually use it thoughtfully rather than flipping every workflow straight to end-to-end the first week because that's the exciting option.

Who is this package actually for?

The honest audience is the business running on one person's judgement across every function at once: the owner who is simultaneously the bookkeeper, the marketer, the person chasing overdue invoices and the one drafting the proposal for next week's biggest prospect, with no realistic path to hiring a specialist for any of those roles soon.

That person has been the least served by AI tooling all year, not because the tools weren't capable enough, but because using them properly required a kind of technical setup time that person has never once had spare.

For that specific person, a package that plugs into the software already running, comes with the guardrail already built rather than something to construct from scratch, and covers the actual functional spread of a small operation rather than one department at a time, is a materially different proposition from another AI tool asking to be learned from zero. It's not a promise that adopting it fixes everything overnight, and nothing in this newsletter is claiming that, in the same way no single tool has ever been the whole answer to how a business runs. It's a description of a gap in what's been available narrowing, for exactly the businesses that have been most stuck on the wrong side of it, for reasons that were never about their own capability or willingness to learn.

What should you actually do with this?

Pick one function, the one currently eating the most time for the least genuine judgement, and see what a pre-built workflow for that specific function looks like connected to the software you already use. Start with the plan-first setting, watch what it produces for a couple of weeks, and only move to end-to-end once you have seen enough of its judgement to trust it with your own.

If your business has been staying at the chat window, not from lack of trying but because every path past it looked like a project with no spare hours to give it, this is worth an actual look rather than a mental note to check out eventually.

The choice of which function to start with matters more than it might seem. Pick the one that's been costing the most in hours rather than the one that feels the most exciting to automate, because the point of this exercise is relief for the person who's been carrying it, not a demo you'd show someone else. The bookkeeping that eats a Sunday night every month is a better first choice than the marketing campaign that's merely tedious, because the Sunday night is the hour that actually comes back to you once the workflow is trusted, and that's the whole reason any of this is worth doing in the first place.

How do you find out where your business stands?

Whether your business is still stuck at the chat window, or has actually moved past it into the kind of connected, workflow-driven use described here, is exactly what the AI Readiness Assessment shows you in about ten minutes, with a result specific to your business rather than a generic score. Click here now to take it.

My book, The Human Edge is launching early Q4 2026. To be first to know when it's available, connect with me on LinkedIn and comment or DM me there.