Services are the new software – but for whom? AI’s $149bn SMB market

SMBs have always been locked out of professional services.

Imagine you run a 20-person construction firm. You’re turning over roughly £4 million a year. You have two or three admin staff handling the back office. Every month there are tax returns to file, VAT submissions to make, invoices to chase, payroll to run for a mix of employed and subcontracted workers, health and safety paperwork for every project, and the occasional legal dispute when a contractor decides your payment can wait another 90 days.

Nobody in the building is strictly qualified for any of it. Five years ago, you tried to get professional help. £100 per hour from a local firm, no promise of how many hours the work would take, and a system you’d need weeks to learn. No time or budget to shop around. So you muddled through – a mix of spreadsheets, gut feel and hoping nothing went wrong.

This is the very real situation for millions of small businesses. There are 33 million businesses with fewer than 50 employees across the UK, EU and US. Of those, roughly 7 million have between 5 and 49 staff – large enough to have real compliance obligations, too small to have ever hired a lawyer or HR team. Below five, the business is often a sole trader with minimal needs. Above 50, firms start to build internal functions. It’s this middle band where the gap is widest. These firms currently spend around $58 billion a year on professional services – but that number is constrained by what they can access and afford today. If you price accounting at $200 a month instead of $2,000 a quarter; if you offer legal compliance as an always-on subscription instead of a £350/hr retainer; if you make payroll and HR one platform instead of three separate providers – we estimate that the same 7 million firms would spend $149 billion. (See here for full market sizing and sources)

Read on for:

• Why 33 million small businesses have been shut out of professional services – and the $149 billion market that opens up when they’re not.
• The four things professional services bundles together – and how AI is unbundling it all
• What Sequoia missed
• Our framework for founders on where to build, scored across every industry and function
• The companies already proving it out, the moats that make them defensible, where we think this goes – and what we’re backing at Emerge.

* * *

Three core reasons have kept professional services firms from serving the SMB market: unit economics don’t work at scale, professional services firms don’t want these clients and SMBs lack procurement infrastructure.

The result is an enormous market that’s massively underserved.

AI has broken the cost constraint

The marginal cost of delivering a professional service has always been a human’s time. Five hours of junior accountant time preparing a set of monthly accounts. One hour of partner time reviewing and signing off. AI changes this. The marginal cost is now constrained to compute, and the cost of AI compute has fallen by roughly 1,000x between 2023 and 2026 for equivalent quality. What cost $30 per million tokens in early 2023 now costs fractions of a cent, and the trend is expected to continue.

What does this mean in real terms? A standard bookkeeping reconciliation that might have taken a junior accountant two hours (billed at £60–£150 an hour) can now be handled by an AI agent for pennies. The economics of servicing a 20-person construction firm have changed fundamentally.

Meanwhile, the incumbents are feeling the pressure from both directions. The professional services pyramid – many juniors supporting a few partners, revenue driven by billable hours multiplied by rate – is built on utilisation. Keep the juniors billing and the model works. However, average billable utilisation across the industry has fallen for three consecutive years to 68.9%, well below the ~85% firms need for healthy margins. Demand for professional services is rising, but the firms can’t convert it efficiently. The pyramid is too expensive to maintain and too rigid to adapt. New firms are being built on AI from day one, and their constraint is compute, not headcount.

Professional services is a bundle which AI pulls apart

Professional services have always bundled four things together:

1. Access to expertise – knowing what the rules are, what the precedents say, what good practice looks like.
2. Application of that expertise to a specific problem – taking the knowledge and applying it to your particular set of accounts, your specific contract, your compliance requirements.
3. Engagement management – scoping the work, scheduling it, communicating with the client, invoicing, tracking deliverables.
4. Accountability – a qualified professional standing behind the output, taking responsibility for its correctness, carrying professional indemnity insurance.

Here’s how AI breaks this bundle apart:

The first three layers fall in cost. The fourth, accountability, stays human. Whether it be signing off on audited accounts or advising on a niche regulatory situation. This is the 10% that can’t be automated. However, this 10% unlocks the automation of the other 90% of the service.

“The predominant capability needed in a post-AI world is discernment — the ability to know if AI output is good enough.”

Chris Benjaminsen, founder, FRVR

This is what creates such a powerful why-now for this space. For the first time, the bundle can be disaggregated and rebuilt at a cost structure that works for a 20-person construction firm, not just a 2,000+ person enterprise.

What the Sequoia piece didn’t cover

Sequoia’s recent article “Services: The New Software” makes the case that the next trillion-dollar company will sell the work, not the tool. Their framework maps every services vertical on an intelligence-to-judgement spectrum: the more a task relies on pattern recognition rather than experience and taste, the sooner an autopilot can replace the human doing it.

What it doesn’t address is the buyer. Sequoia’s examples – replacing an outsourcing contract, substituting a budget line, doing a vendor swap – all assume a company with procurement infrastructure, an existing services spend and someone in-seat to manage the transition. That describes an enterprise. It doesn’t describe the millions of businesses who never had access to these services in the first place.

But the autopilot thesis isn’t just about doing existing work cheaper. For SMBs, it’s about unlocking work that was never economically viable to begin with.

The first wave of AI in professional services was co-pilots – tools that make existing lawyers, accountants and compliance teams faster. Harvey for legal research, Vanta for compliance workflows, Rippling for HR management. That made sense: enterprise budgets are there, the buyer is obvious and the professional is already in-seat ready to use the tool.

Co-pilots don’t work for SMBs as there’s nobody to operate them. A 20-person construction firm doesn’t have a lawyer to make faster with Harvey, or a compliance team to arm with Vanta, or an HR person to supercharge with Rippling.

In comes the autopilot, providing the full service itself, end-to-end. Not a tool that helps a professional do their job, but the replacement for the missing professional entirely. Perfectly suited to SMBs with no dedicated headcount for that specific function.

Building an autopilot is genuinely hard, which is why so few exist. You need domain expertise deep enough to handle the full workflow. You need to earn trust from a buyer who has never used the service before, let alone an AI-powered version of it. And you need to get pricing right for a customer who has no frame of reference for what the service should cost.

At Emerge, we think the autopilot category is where the defining companies of the professional services space get built over the next few years.

SMB services are a fundamentally different product

Selling to SMBs is nothing like selling to the enterprise. In practice, it’s much closer to a B2C motion than a B2B one. The buyer is a single person, the decision is fast and the tolerance for friction is near zero.

This creates a set of go-to-market realities that founders in this space need to be prepared for:

No specialist buyer. The CEO buys everything. They can’t evaluate hourly rates or know whether a task should take two hours or six. Outcome-based pricing fixes this: “£50 a month for all your employment contracts” is a decision a founder can make quickly.

No tolerance for integration work. There’s generally no IT team, so the product must work on day one. This also means bundling related functions makes sense – accounting plus tax plus payroll in one place – so systems talk to each other and you’re not reconciling data across three different platforms.

Risk transfer matters more than features. An SMB founder handing their accounts to an AI company they’ve never heard of isn’t just outsourcing work; they’re also outsourcing risk. They don’t have internal expertise to spot if the output is wrong. So they’re thinking one question: “if this gets it wrong, who’s liable?” This is why the certified human layer matters. A chartered accountant signing off on the accounts or a solicitor reviewing the contract is a risk transfer mechanism. Almost all of the autopilots we’ve seen or spoken to have humans in the loop. Chartered accountants signing off, lawyers reviewing, specialists handling edge cases. The delivery model is AI handling 90% of the volume, with one human scaling to 50–100x the clients they could serve manually. AI helps one professional scale like a full department.

Founders: if you build for a buyer who has no time, lack of relevant expertise and little tolerance for complexity you’ll build something they never want to leave.

Where to build: our framework for founders

For founders starting out in this space, choosing a niche and building from there makes earning trust with CEO buyers significantly easier. It gives you a focused way to grow in the early stages.

At Emerge, we see three tensions determining which sector/niche is exciting to build in:

Is the spend compulsory or discretionary? Compulsory means the SMB has to do this whether they want to or not: tax filing, employment contracts, safety compliance. Discretionary means they can choose not to: marketing, strategy, financial planning. We prefer compulsory. You don’t need to convince the SMB they have a problem as the regulation already does that. CAC is lower and retention is stickier.

How high are the stakes if the AI gets it wrong? Low stakes means easier to automate fully: marketing content, basic bookkeeping, lead generation. High stakes means the delivery model needs certified professionals in the loop: tax filings that will be audited, contracts that create legal liability, compliance certificates that carry penalties. This adds cost and complexity to the build. But high stakes is also where the defensibility is strongest: if it’s hard to get right, it’s hard for someone else to replicate. We’re most excited about founders deliberately choosing high-stakes categories and engineering the human layer to make the economics work.

“Where the stakes are massively high, or the target market is a small number,  human experience is still super useful.”

Charlie Schilling, CEO, Macabacus and Emerge Venture Partner

Is the workflow repeatable or bespoke? Repeatable means the same basic process for every client: monthly bookkeeping, payroll runs, standard contract review, safety audits. Bespoke means every engagement is different: M&A advisory, complex litigation, turnaround consulting. The best autopilots are built on repeatable workflows that handle 90% of the work, with the last 10% of bespoke variation handled by the human layer. That bespoke tail is what makes the service defensible; it’s the part competitors can’t easily replicate with a generic model.

We mapped out every industry’s spend on professional services and then scored every industry-by-professional services cell against the above three tensions.

The highest-conviction cells – all scoring 80 or above out of 100 – are dominated by accounting and compliance in physical industries:

Looking more closely at the trends we are seeing:

• Accounting is high conviction in every industry. It’s compulsory, high-stakes and repeatable. Every SMB has to file taxes, run payroll and close their books – and getting it wrong means penalties. If you’re looking for the widest entry point, this is it.
• Compliance scores highest in physical industries – construction, healthcare, manufacturing – where H&S, environmental and sector-specific regulations create mandatory, recurring workflows. The penalties for getting it wrong are serious, which is exactly what makes these cells defensible.
• Legal is exciting. Every SMB has legal exposure but most only engage a lawyer when something has already gone wrong; the entry cost has always been too high for preventative work. The stakes are real and the need is continuous. The question is whether per-client volume supports autopilot economics. In construction, where contract disputes are constant and the sums are large, it clearly does. In lighter-touch industries, founders need enough recurring legal surface area to make the unit economics work.
• Sales and GTM scores lowest across the board: it’s discretionary, low-stakes, relationship-based and resistant to productisation. Marketing sits in between: automatable and repeatable, but discretionary spend and low stakes mean it’s easy to enter and hard to defend.
• HR and people sits middle across most industries. Compulsory payroll requirements and industry-specific regulations keep it relevant, but volume is limited in <50 person SMBs.

For the full heatmap scoring, see here.

The companies proving the thesis

This space is still extremely early with no clear breakouts. The market has been focused on co-pilots to date, which makes sense – co-pilots are easier to build. They’re essentially a collection of point solutions that a human uses. Autopilots are harder because they have to autonomously run end-to-end processes, not just assist with individual steps. 

Looking specifically at autopilots, we’ve put together the below market map. As you can see, most of the companies we identified are specifically focused on SMBs. This reflects our view that at this stage, the autopilot category is extremely well suited to small businesses.

Let’s look more closely at a few of the companies in this space:

“Most people don’t realise how hungry SMBs are for AI-powered professional services. We could barely keep up with demand.”

 

Joshua Spiegler, founder, Glorya

The moats that stick

Tacit domain knowledge. Every vertical has unwritten rules – which contract clauses actually get disputed in construction, which compliance checks inspectors care about most, which tax structures HMRC flags for specific professions. This knowledge doesn’t exist in any training set. It can’t be prompted into a frontier model or configured by an internal team. It has to be reverse-engineered by teams who understand both the domain and the technology, and it accumulates through thousands of client engagements. The longer you operate in a vertical, the more of this tacit knowledge your system absorbs, and the harder it becomes to compete with.

Proprietary cross-client data. An AI accounting firm serving 1,000 therapists learns patterns that make it better for therapist number 1,001: which deductions are common, what tax structures work, where audits flag issues. This is data generated by the product itself – it didn’t exist before the workflow created it. Horizontal players get less of this advantage because the patterns don’t transfer 1:1 across industries.

Regulatory infrastructure. This is less a moat and more a barrier to entry, but still meaningful. In Europe, offering regulated professional services requires specific licences and insurance depending on the jurisdiction. In the UK, legal services need Solicitors Regulation Authority authorisation. In Germany, tax advisory requires a Steuerberater licence. Across the EU, professional indemnity insurance is standard. Getting these takes time, money and specialist knowledge – a new entrant can’t just spin up an LLM and start filing tax returns. It doesn’t compound like data does, but it buys time and credibility while the real moats are being built.

The path to scale follows naturally: build the domain knowledge and data moats early, then expand horizontally into adjacent services from a position where both the expertise and the compliance groundwork are already done.

Where this goes next

This category is still extremely nascent. Here’s how we expect it to develop:

1. Co-pilot to autopilot convergence. Today’s co-pilots will push towards becoming tomorrow’s autopilots as models improve. But co-pilots face a real tension. Harvey’s customers include law firms and in-house legal teams. If Harvey automates the full legal workflow end-to-end, the law firm piece of the business starts cannibalising itself – the more it automates, the less the firm needs to exist. The in-house teams have more room, but the product is still built for someone who’s already there. Purpose-built autopilots don’t carry that baggage. They’re designed from day one for the case where no professional exists, which is most of the SMB market.
2. Autopilot growth path. The autopilot starts as the missing professional for a 20-person firm. The firm grows to 80 employees and hires a head of finance. The autopilot becomes the team underneath her – one senior hire managing a fleet of AI agents rather than five junior hires. This is a retention mechanism, not just an acquisition story.
3. Horizontal expansion. An AI accounting firm that owns the financial data becomes the natural provider of tax, payroll and CFO services. Whether one platform can become the full AI back-office is the trillion-dollar question. Our instinct is that it’s extremely difficult to be genuinely excellent at everything. The more likely outcome is a handful of vertical-first companies that expand into adjacent functions, rather than one horizontal platform that does it all.

What we’re most excited about at Emerge

We think AI-native professional services for SMBs is one of the most compelling investment categories of the next decade. Here’s what we’re looking for:

“When the value of software drops to zero, the value of service goes up. We’re looking at a renaissance of professional services for SMBs.”

 

Joshua Wohle, founder, SuperAwesome & Mindstone, and Emerge VP