What Comes After Selling Work?
Writing about how technology and people intersect. By day, I’m building Daybreak to partner with early-stage founders. By night, I’m writing Digital Native about market trends and startup opportunities.
If you haven’t subscribed, join 70,000+ weekly readers by subscribing here:
SubscribedHey Everyone,
After 6+ years of consistent weekly pieces, I took a bit of a breather over the summer. We launched Daybreak Fund II and things have been more than a little busy. I want to make sure Digital Native continues, but we’ll shift away from a weekly cadence and I’ll instead write more sporadically, whenever I have something (hopefully) interesting + additive to say. I’ll also be writing more with my partner at Daybreak, Jared Newman.
We wrote this piece together, refreshing our long-held thesis in AI-native services. This piece touches on what we’ve been observing in the markets and opportunities we see for “selling the network.”
Let’s get into it…
What Comes After Selling Work?
If you’re living and breathing early-stage startups day in and day out, it might seem like we’re in the middle innings of AI-native services (AINS).
We’ve seen new kinds of services firms built in just about every vertical: roll-ups, JVs, neofirms, holding companies, and good, old-fashioned AI-native startups tackling niches within law, finance, travel, insurance, compliance, advertising, and just about everything else. Some of these firms already have real proofpoints of success.
Take Crosby, which is a good marker of where we are: after 18 months of reviewing contracts, Crosby attorneys have generated enough data for the eng team to fine-tune an open-weight model that meaningfully predicts which AI redlines a human lawyer will accept. This is an early but meaningful step toward the “AI-native” vision for services that we wrote about in Digital Native two-ish years ago. America has shifted from an agriculture economy, to a manufacturing economy, to a services economy; and now, AI is eating services.
When it comes to AINS, VCs like to point to one stat in particular: for every $1 businesses spend on software, they spend roughly $6 on services. When you think about it this way, which has the bigger TAM: the company selling technology to law firms, or the company that becomes the law firm itself? (In other words: Harvey/Legora vs. Crosby?) The latter may have more to prove by way of margins (maybe!) but it should theoretically have higher upside. The TAM is less bounded.
When you think about the services yet to be consumed by AI, you realize we’re still early innings for the AINS thesis. We’re also beginning to see real risks surface. Here are two we’ve been paying attention to:
Large, incumbent service providers have not all been slow to adopt AI internally. The canonical example here is Kirkland & Ellis, the world’s highest grossing law firm (until it announced recently that it would stop disclosing financials). Kirkland is investing $500M into its own, proprietary AI stack. This sum is surplus to what they already spend on tools like Harvey. But the interesting question is not whether Kirkland will build or buy its technology, but rather whether these big-name brands will actually be effective competitors over time. And if Crosby is right that the moat is the judgment data that lawyers generate when reviewing AI output, Kirkland, with thousands of lawyers, may have far more of it. When you look around, you seeing Goliaths that aren’t caught flat-footed. Just recently, Goodwin announced its own stack. Outside of law, McKinsey has Lilli, EY has spent $1.4B building EY.ai, and PwC is one of OpenAI’s largest enterprise customers. The Innovator’s Dilemma always looms large, but what’s different today is just how much money and talent the frontier labs are dedicating to legacy services firms (see: Ode and OpenAI Deploy Co).
Selling services is hard. It’s often harder than selling software. Ask any rainmaking partner what her job really entails. Ask any insurance broker how much golf is required to upsell a 10-year customer. Services sales are, fundamentally, relationship-based (read: slow). There’s often no way around barriers to speed and adoption. We’ve seen some AI-native services startups hack distribution by hiring for it, poaching producers with extensive books of business as a way to fill pipelines. It’s a tricky playbook, given that you want value to accrue to the firm brand, not just to the salesperson. Salespeople who, sometimes, are bound by non-solicitations…
Sell the Network
We’ve also noticed a different response to both of these problems. A handful of startups are beginning to ask whether the AI-native service itself needs to be the only sale.
It’s a slightly ironic conclusion to reach at the precise moment AI-native services are starting to work. But consider the two risks above. If a service is the only thing you sell, you have to compete with incumbents who already possess the customers, the brand, and potentially even more proprietary data than you do. And, assuming you build something better, you still have to convince those customers to buy it, often one relationship at a time.
So what if, instead of charging for a service, you give it away for free?
You might be thinking, hold on… this isn’t software. You can’t just give away services for free! But if models continue improving, some of the work (and yes, even some of the outcomes) may eventually become cheap enough that the better business is to give them away. Better, how? Well, services sit upstream of something much more valuable: a network of transactions, suppliers, customers, capital, or data that may be where the real money lies in the long term. This is something incumbents can’t easily copy. Kirkland can build a better AI stack, but it can’t give redlines away for free.
Our portfolio company Rockwell offers a useful early example. The founders are building an AI-native benefits department for mid-market employers. We’ve long been excited about investing in founders tackling one of the biggest challenges in the US today: rising healthcare costs. The majority of Americans (about 160M) get health insurance through their employer. The yearly price for employees to cover their family is now $7,471 a year, +9% from 2025, the first time prices have increased more than 7% three years running. This will only worsen next year: for 2027, employers may be facing the biggest health-insurance increase in at least twenty years.
Rockwell isn’t just a standard broker, and it’s not just slapping the “AI-enabled” moniker on top either. When employers partner with Rockwell, their employees get a benefits and care concierge that helps them figure out where to go, what’s covered under their plan, and how to navigate care. All for free. That interaction sits directly upstream of enormous healthcare spend and data. So rather than ignore the immense value post-benefits, the way an old-school broker might, Rockwell can use its care concierge to aggregate demand and steer members toward the providers, pharmacy programs, care-management offerings, and other services it increasingly controls or contracts around. The free concierge becomes the engagement layer; the real business is the network and healthcare spend flowing beneath it. That’s much more than a traditional broker could ever offer.
Another example can be found in personal injury, where we’ve also spent lots of time. Today, PI firms buy software and services for intake, case management, treatment coordination, medical records, demand generation, and so on. An AI-native services company might bundle that work together and charge the firm for a better outcome. But what if someone gave all this pre-lit work away for free? Then, a free case-management provider could see when a plaintiff needs diagnostics, PT, or medical treatment and route that person to a vetted provider network. If those providers want the patient volume badly enough, they may pay for that lead. Suddenly the pre-lit isn’t the product at all. It’s just a mechanism for aggregating the more valuable network.
In the past, Google made search free because the network of commercial intent data was more valuable than search itself. Technological abundance rarely destroys value; rather, it relocates it, pointing out what was actually scarce all along. Services firms have always sold time, but what if time is no longer the only asset we have to sell?
So what are we looking for?
At Daybreak, we’re still excited about investing across AI-native services. But we’re increasingly looking for companies that innovate on technology and on business model. Sometimes this means giving away services for free, but there are plenty of other compelling ways to monetize some combination of risk, data, or capital, rather than time alone.
Crosby, of course, pioneered this way of thinking by ditching billable hours for fixed, per-contract pricings. Clearly, it’s a move that’s resonating given their recent OOH ad campaign blanketing the New York subways:
We’ve seen other novel and creative business models as well:
Risk-bearing models — If a firm gets good enough to predict outcomes, it might be more lucrative to price for uncertainty around those outcomes. For instance, a compliance firm could insure against certain kinds of failures, rather than for a SOC II verification. Suddenly, you’re selling a kind of insurance, which is famously among the most profitable business models of all time.
Financing — Shopify and Toast proved that lending can be very lucrative, so long as you have true visibility into the right kind of creditworthiness data. Services firms often sift through exactly that: invoices, claims, contracts, settlement values, and vendor notes. The easiest example here might be an AI-native outsourced bookkeeper, who knows a client’s cash needs before a bank does.
In the AI revolution, the future is arriving faster than ever. What once took years now takes months now takes weeks, and it’s this acceleration that gives us confidence that selling work might soon go the way of selling seats. We already know how to use AI to deliver old services more efficiently. Now, the question becomes: what entirely new business models can be built on top of those old services?
Credit to Jared Newman for really driving forward this piece and much of our thinking on this category.
Thanks for reading! Subscribe here to receive Digital Native in your inbox each week:
© 2026 Rex Woodbury
548 Market Street PMB 72296, San Francisco, CA 94104