On 31 March, Wing Venture Capital and the Newcomer team published the eighth annual Enterprise Tech 30. Ninety-eight investors holding $2.6 trillion in collective assets agreed on sixty companies. Sixty out of more than fifteen thousand in the running. The closest thing the enterprise software industry has to a map.
I read it on the day it dropped. By the time I got to the bottom of the list, I knew I was on the right side of two of three bets it blessed.
The three bets, as named by Wing partner Peter Wagner: voice AI, vertical applications, and services-as-software. Eighteen months ago, two of these were what Wagner now openly calls "killing fields." Investors believed verticals had no addressable market. They believed services would never get software margins. Voice AI existed but it was mostly speech-to-text APIs and call-centre analytics, not products in the wild.
Then, very quietly, all three got blessed.
ElevenLabs took the number one spot in Late stage. Granola took number two in Mid. Three more voice companies showed up across the early and mid stages: Wispr Flow, Retell AI, David AI. Five voice companies on a list that picks one in every two hundred and fifty contenders.
The verticals: Basis (accounting), Avoca (home services), Pace (insurance ops), Assort Health, Crosby (legal, structured as an AI law firm rather than a software vendor). Wagner's own words: "Investors have suddenly opened their hearts to vertical applications." Two years ago, the same investors called those companies unfundable.
This is the part where I admit my hand.
I have four voice agents in market
Buggy Smart calls London pubs to find out which ones welcome buggies (American: strollers). It has now done thirteen thousand four hundred and seventy-two conversations. Nine hundred and thirty one venues classified. The agent has a voice, a name (Poppy), and a brand. Pubs answer the phone, talk to her like she's a real customer, and then we publish the result.
First Order calls London restaurants and asks for the chef's pick. The agent gets a yes about one time in seven. The other six times the host laughs, transfers her to nobody, or hangs up. Both numbers are useful.
Moshi answers the phone for restaurants that don't have a person to do it. UK-only, restaurant-only, cheap, fast. Built it because UK restaurants miss calls during service hours and that is revenue dropping on the floor.
Queue Index scores how long the queue is at the venues that matter to Londoners. It calls during peak hours. Sixty-nine percent success rate in the first week.
That is not the brag. The brag is: I shipped them all before ElevenLabs hit eleven billion dollars. They use ElevenLabs underneath. So when someone asks why I think voice AI is the breakout vertical of 2026, the answer is in market, talking to humans, in three London accents.
And a portfolio of vertical apps
Oishii is a Japanese restaurant directory for London. Eighty-four restaurants, every one verified by hand. People who want sushi do not want a restaurant directory; they want sushi recommendations. Specific.
Visible Shelf is a 3D bookshelf you can browse. Two hundred and ninety-five books. Every one has a take, an audio reading, a category. It is not Goodreads. Goodreads is horizontal. Visible Shelf is mine.
London Pub Guide is a thousand pubs of London with neighbourhood guides, queue advice, beer takes. Pub guide London is what people actually search for; that is the URL.
Little London is for parents in north London with small children. Where to take a kid who is two. Where to take a baby. Where to go on a Saturday morning when nothing else is open. Specific to a postcode, specific to a stage, specific to a season.
Each one is laughably narrow. That was the point. Wagner just told you why narrow won.
The thing nobody tells you about killing fields
There is a line in the deck I made out of this list that I keep coming back to. Three bets that Wing's Wagner called killing fields just got blessed.
Killing fields. Verticals were killing fields because the addressable market looked too small. Services were killing fields because the gross margin was too low. Voice was killing fields because the technology could not feel human.
All three of those problems got solved while VCs were not looking. Verticals got bigger because AI gives a vertical product the ability to handle ten times the workflow it could before. Services compressed toward software margins because the labour curve broke. Voice crossed the human threshold because the model layer caught up with what humans expect.
I was building in two of those three because the killing-field reasoning was already wrong. You can see when a market is being mispriced if you spend long enough in it. Mostly you cannot get a VC interested in talking about it; their pattern matchers are calibrated to the last cycle. So you build anyway.
What this means for the next twelve months
This is the part of the post where I am meant to give you something you can use.
If you are building, the rule for the next twelve months is: build the killing field nobody wants to fund. The list has shown that the past killing fields are the present gold rush. The current killing fields are the next ones.
What is the current killing field? My nominee: AI products that refuse to perform. The category nobody is funding right now is products that are deliberately quiet, opinionated, slow, and expensive. The opposite of the agentic supermarket. There is a list waiting to be made of products that wait, that do not reply on demand, that defer, that say no. That is eighteen months out from being a Wagner quote.
Or maybe it is something else. The point is the same. You cannot time institutional consensus. Build the thing you would want to use, the thing that is so specific it embarrasses you to describe at a dinner party, the thing where the addressable market fits in your contacts list. Then keep doing it for two years.
One more observation
Newcomer flagged a sleeper in the analysis. Decagon (#4 Late) beat Sierra (#6 Late) in customer support. Sierra has the higher valuation. Decagon won the VC vote anyway. Sentiment is leading valuation by two ranks.
That happens when builders inside a category know something the market has not priced yet. The next round usually corrects.
I take that as a permission slip. Build what you would back if you were on the inside. The list comes later.
If you want the longer version, the full talk is here: 22 slides on what the list says, what got dropped, why services-as-software just got fundable, and the verdict on which bet to double down on next.
