Emergence Capital's $150 Million Bet Isn't About AI

Emergence Capital has put more than 150 million dollars into a handful of companies selling something the venture industry has decided to call AI-native services. The bet is specific. These firms are going after roughly 200 billion dollars of the Big Four's annual professional-services revenue, work that used to be billed by the hour and is now being sold as a priced outcome instead.

I have read the pitch decks. I have read Emergence's own writing on why they think this works. And the thing that strikes me is not the AI. It is the staffing model underneath it.

Every AI-native services company Emergence backs does the same thing. It hires the domain expert first, the lawyer, the accountant, the insurance broker. Then it pairs that person with an engineer who understands the technology. The company gets paid only when the outcome lands, not for the hours either of them spent getting there. That is not a technology story. That is an organisational design story wearing a technology coat.

It is also the fastest way to tell a genuinely AI-native supplier from one that has re-labelled the old workflow. Look at who they hired, what activities disappeared, and how they price. If the staffing model and the invoice look like 2019, the AI is in the pitch, not the production line.

I have seen this pattern work before venture capital gave it a name.

Some years ago I led an AI-enabled operating model rebuild for a global pharmaceutical client. The starting point had nothing to do with a model or a licence. It was a workflow with no baseline. Across twelve major markets, accounting for roughly 82 percent of the client's global business, plus a long tail running over forty more, the organisation was running more than a thousand separate content agencies. Content reuse across that network sat at 6 percent. Around 100 million dollars a year in content-creation spend could not be tied to any specific outcome.

We did not start with a platform. We started with the workflow, how the work actually got done, market by market, and why it differed so much between them. Only once that was mapped did we redesign it, consolidate production, and put a single system underneath it, with automation increasing over successive years.

We costed it activity by activity, the same Activity Based Costing discipline I first applied to content production at Tag twenty-five years ago and later built into Hogarth's ways of working. Once every activity has a price, the conversation about what to drop, automate or keep becomes a conversation about numbers.

The result was a reduction from over a thousand agencies to five global entities organised by therapeutic area. Content reuse rose from 6 percent to 70 percent. The programme delivered more than 11 million dollars in direct savings and more than 15 million dollars in cost avoidance, money freed to be spent on work that actually moved the needle rather than work that simply existed.

None of that came from a better model. It came from redesigning how the work was done before any technology was chosen, and from treating the people closest to the work, market by market, as the source of the design rather than the recipients of it.

That last part is easy to miss, and it is the part venture capital's own model quietly confirms. The AI-native services companies Emergence backs are deliberately going after work that is rules-based and judgement-light, bookkeeping, claims processing, renewals, the long tail of legal and financial work that has always run on volume rather than expertise. That is precisely where a machine can take the reading. It is not where deep, holistic domain expertise stops mattering. It is where it starts to matter more, because someone still has to decide what the machine should be doing, and get the humans whose work is changing to actually believe in the redesign.

Five percent of custom enterprise AI tools reach production. Ninety-five in a hundred do not. The gap is rarely the model. It is the absence of a baseline, an owner of the measure, and a workflow that was ever actually redesigned rather than automated in place.

Emergence's own 200 billion dollar target sits inside a global services market estimated at 4.6 trillion dollars. Pharma and life sciences are a meaningful share of that market and are, on the adoption curve I use, still mostly at stage one or two of four. The organisations that move first on redesigning the workflow, not on buying the next licence, are the ones who will own that ground before someone else prices it as an outcome and sells it to them.

The hype cycle around AI was never wrong about the scale of the shift. It was wrong about where the shift actually happens. It is not in the model. It is in how humans and the systems around them are asked to work, which has always been the real production line, long before anyone called it artificial intelligence.

That is the work. Not what to buy. How to run it.