Using AI only to cut costs is a pretty sad thing
Netic founder Melisa Tokmak says private equity's first meeting with her always starts with cost cutting, but what she actually sells is net new revenue for these companies — her customers have already made $600 million more because of it.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Answering a phone call is much harder than the model
Netic handles scenarios like "it's -20 degrees and the heat is broken": a customer finds a service provider through search, an aggregator or an LLM, then calls, texts or goes to the website, and the one answering is Netic's agent. The hard part isn't speech recognition, it's the reasoning chain behind it — what kind of house you live in, whether there's a history, whether to send someone today or tomorrow, whether to send the old-timer who can only fix boilers or the younger one who can handle new systems, whether this customer's lifetime value justifies priority dispatch. Melisa says these operational requirements are "very complex," not as simple as "Eli's heat is broken, Melissa goes to fix it."
— Melisa TokmakThe pain point in this business is that nobody answers the phone at 6am
She describes the daily reality of a billion-dollar-revenue company: the business starts at four or five in the morning, but at that hour there's not a single person in the company, and technicians don't start until six. By the time the manager arrives, three people may have quit and five haven't shown up. Meanwhile customers start flooding in from six in the morning — because there's a heat wave across the country. These businesses are cyclical: winter, summer, holidays, the season when people start working out, and the value has to be captured in those few months to support yourself and your employees. So Netic first entered through "answering overflow calls," but today more than 70% of its customers are AI first, and every one of their customers' first contact happens on a Netic agent.
— Melisa TokmakShe doesn't do roll-ups, because that isn't her skill
Faced with the question "why not just buy up a bunch of companies and optimize them with AI," she gives three reasons. First, the most important thing in a roll-up is M&A itself, and she isn't an M&A person — she's an engineer and a product person — and she doesn't want to do a business where "what I can offer isn't the most important thing." Second, in a roll-up the product you build for the company you just bought can't really be reused for other companies, which amounts to locking yourself into a handful of industries. Third, what she wants is for "every real-world business to run on Netic" — letting these companies focus on what's genuinely differentiated, namely labor and quality of service, and leaving the rest to Netic.
— Melisa TokmakFor robots to take over, every building would first have to be torn down and rebuilt
She compares AI for the real world to different chapters of the same book: the robot chapter is in the future, but in the industries she serves it's still quite far off. The reason is the non-standardization of the physical world — if robots were to do what these companies do today, every building outside the window would first have to be 3D-printed or thoroughly standardized, and she doesn't see that happening. Today's robot capabilities are still far off in dexterity: different kinds of screws, differently structured houses, how to climb up, how to squeeze into tight spaces — often you have to open up an entire wall before you know what needs fixing. Add the human factor — these moments are often the worst day of a customer's life, their home is flooded, or they just want to relieve stress by playing tennis.
— Melisa TokmakThe big labs aren't competitors, they're partners
She thinks this question was "can Google do it" ten years ago, and now it's "can the labs do it." Her judgment: on core capabilities the labs genuinely can, but there are things they won't invest in. She gives two specific points — OpenAI ships products extremely fast, but "kills products extremely fast too," and enterprises in these industries don't want that cadence; Anthropic leads on coding agents because of focus, but in the enterprise context what you see is about 20 products and you can't tell what's actually happening. She also says lab researchers care about the most generalizable solution, which in this context amounts to "wait for AGI and then ask how to solve essential services," which she considers operationally and intellectually "a bit lazy."
— Melisa TokmakFounders are now too afraid of the labs
The host observes that four years ago he was investing in vertical applications like Harvey, Perplexity and Decagon, whereas now many founders worry too much about the labs' roadmaps and so don't dare enter new verticals — in an earlier era people would have fought a round first and asked questions later. Melisa's response: that's because many people now build things thinking "how do I exit immediately," rather than treating founding as something you commit decades to. She says that before starting her company she seriously looked for a job — first asking whether there was anything she wanted to build, then whether there was anything else she was interested in, and finally whether there was anyone whose mission she wanted to advance. She also calls out the "permanent underclass" mindset popular among Gen Z: if you haven't made money within 18 months and haven't learned everything within 6 months, you'll be poor forever.
— Melisa TokmakScreening for agency means seeing whether you stuck with it
Her way of screening for agency isn't asking one question, it's digging into life experience: if you're a new grad, what did you do in college? What did you do to get into college? Is there a project you genuinely cared about? The key isn't "I did this over the weekend," it's whether you kept doing it. One question she often asks is "what's the hardest thing you've ever done in your life" — the answer can come from anywhere, it may not even be something you chose; what matters is how you reacted, what you controlled, and how you turned the part you could control into a different future. She gives an example of someone she just hired: the person said their life is very simple, they care about work, they have a crazy discipline around health, and the hardest thing was keeping that routine for over 15 years without getting bored. She says that's a very creative answer.
— Melisa TokmakThese industries aren't old school, they're primitive and tech-forward at once
She thinks seeing these industries as old school is a big misunderstanding, and that the most tech-forward, most business-focused people she's met are precisely in this industry. She gives a number: a deal just signed might be a $500,000 contract, from start to finish in 14 days — not because AI is a magic potion, but because they examine very carefully whether the value really exists. She gives the example of roofing companies: these people were already going to hire people to knock on doors, and now Netic plugs in satellite data to judge how hurricanes affected roofs in different blocks and what materials to use, automatically feeding that into the agent's context, so it can both converse better with customers and identify who to go after.
— Melisa TokmakPrivate equity's playbook has already changed
She says private equity's old playbook — find a gem with a beautiful multiple, swap out the team, create value, sell it — no longer works, because those undiscovered gems no longer exist, and the playbook has become "how do you create real value in these businesses." She observes two extremes: some go too deep, treating AI as software and expecting results in the first week; others move too slowly. She thinks the right posture is to separate AI from vaporware, look at real ROI, and look at whether results keep getting better rather than worse over a full year. She also mentions that private equity now hires AI operating partners, or keeps a few AI-oriented engineers to do internal education.
— Melisa TokmakThe first meeting always starts with cost cutting
She says private equity's first round of conversation always focuses on cutting costs, because they've never seen a product like Netic, while what she really wants to talk about is net new revenue. She thinks this requires proactively opening the conversation and showing them concrete examples, otherwise the topic falls back to how to squeeze the bottom line and cut some costs. Her exact words: if AI is only used to cut costs, that would be pretty sad. She also stresses that private equity is more about advising portfolio companies, and can rarely shove things down a company's throat, but it can be a good advisor telling companies what to examine.
— Melisa TokmakIn their own words · checked verbatim
it would be a funny question to these enterprises right open AI builds amazing products really fast but also it kills them really fast
Melisa Tokmak13:16
they really care about solving the most generalizable way of the problem right so in this case maybe looking at the problem we're solving the answer would be well when we get the a AGI we'll ask how to solve it for essential services and I think that is both operationally and intellectually a bit lazy thinking
Melisa Tokmak14:16
the Christian shoe maker, you know, doesn't honor God by putting little crosses on the shoes. It does so by building the best shoe, right? The best shoes. Uh because God cares about craftsmanship.
Melisa Tokmak18:20
we have made so far, I think over $600 million for our customers that have been really generated from AI handled interactions
Melisa Tokmak29:25
It would be pretty sad if we used AI only for cost cutting.
Melisa Tokmak30:26
We keep making it easy and easier easier and easier to make that choice but still I guarantee you majority of the world will not be making that choice.
Melisa Tokmak32:28
Figures
| Share of Netic customers that are AI first | More than 70% | 6:12 |
| Deal size of Meta and Scale | About $28 billion (the host says "around 30") | 7:13 |
| Revenue Netic has created for customers | More than $600 million | 29:25 |
| Age of Netic | Two years | 4:09 |
Glossary
- roll-up
- Buying up a batch of similar companies, merging their operations, then using technology or management to raise the combined value.
- essential services
- Real-world service industries that keep daily life running: HVAC, plumbing, pets, auto, and so on.
- harness
- The software layer outside the model that handles orchestration, context and tool calls, and determines actual results.
- vaporware
- A product that is heavily hyped but cannot actually be delivered or has no real effect.
How to listen
Founders building vertical AI applications who are agonizing over whether to touch "slow industries," and investors watching how AI reshapes private equity portfolio companies.
The company intro at 0:00-3:53 and the personal reflections after 31:14, both noticeably low on information density.