We Built Our Own Salesforce in Months. Here's Why We're Cancelling the $600K Contract | Curative CEO

Summary

• Fred Turner’s company died in December 2019, and the wind-down preceded the $5BN COVID-testing business. He sold Curative’s predecessor’s lab license for $150,000 to pay creditors, then five months later paid $27 million to acquire a Southern California lab for the same license — funded out of forward revenue from a police-and-fire testing contract. “Timing is everything.” Total COVID testing revenue: about $5 billion over three years, 7 to 7,000 employees in nine months, and a peak of 206,000 people tested in a single day (the ASR garbles this as 26,000 in places).
• The operating insight is the tradeable one: Quest and LabCorp are built for 1% efficiency gains, not 10x, so the incumbent’s core competence was the disqualifier. Curative built what Turner sells as an “orthogonal supply chain” — sourcing swabs from electronics vendors, using glass-and-plastic filter plates instead of magnetic beads because those “all come from basically two factories in China.” Outside a pandemic it’s a terrible design; inside one it’s the approach that scales.
• Margins were a surge function, and Turner is blunt that the model was structurally loss-making between waves: “we would lose money on every test during the dips,” with 7,000 people and full fixed cost carried through the lulls because reimbursement was deliberately overpriced to force capacity to exist. The 2.5 million vaccinations were worse — “We also lost a ton of money on that. That was a terrible business” — done because partners wanted them and as “giving back,” at a loss on every dose.
• The AI call, from a real P&L: credentialing went from ~$50 and two-to-three months to 20 cents and 12 hours on a Claude-based agent, and the contracting agent “Gwen” took Curative from ~100 provider contracts a week to ~100 a day at $70 versus $1,500–2,000 per contract. Anthropic spend has 6x’d every month for six or seven months, tens of thousands to millions monthly. Asked if a 2x or 5x price hike would change usage: “It would still work.”
• Back-office SaaS is under attack, and he’s putting numbers on it: Salesforce cancelled ($600K/year, replaced in two months by a vibe-coded internal CRM), a full-time Salesforce admin replaced by one great engineer, an in-house claims system replacing an off-the-shelf platform, and ~80% of total SaaS spend cut this year — tracked on an internal slide of renewal dates and “whose job is it to tell them we’re not renewing.” What survives is infrastructure: Sentry, and Slack because of integration lock-in.
• The competitive thesis against the four incumbent payers is political, not technical: “If you’ve built a company of 100,000 people and in order to get this margin improvement 50,000 of them have to be laid off, somebody’s fiefdom just got a lot smaller. And so they will do it slowly over 10 years.” Meanwhile Curative goes from 650 people to roughly 400, and Turner is bullish on AI-driven earnings precisely in sectors like insurance that have no other growth lever — the 85% MLR floor caps profit at 15%, so admin efficiency is a way to grow earnings without worsening the product.
• Two forward calls worth marking: Anthropic “could be 10 trillion” in three years, and spend on models will run 2–5x total developer salary (versus Benioff’s cited 3.8% today). His nomination for the job that doesn’t exist yet is “agent supervisor” — exception volume scales with the 10x throughput, and “if we ever get a more expensive model and it costs $100 per million tokens you probably wouldn’t use it for the core workflow, but you could maybe use it as a supervisor.”

Deep dive

1. The wind-down preceded COVID scale

• The company that became Shield was killed by a strategic in late 2019: term sheet signed with a large public diagnostics company, three weeks of docs, second round of documents — “and then their CEO killed it because it was too competitive with their core products.” They had told all investors they had the lead. Three weeks of cash left.
• Harry’s question earns the best answer of the segment: would he be where he is if that round had closed? “No. No. Cuz I don’t think we would have pivoted as hard into COVID when COVID hit.”
• The detail that makes the story: in December 2019, as part of the wind-down, he sold the lab license — painstakingly acquired over two years — to a San Diego company for $150,000 to pay creditors. Five months later he paid $27 million to buy a Southern California company for the same license. “Timing is everything.”
• His lesson from the failure is unsentimental: the seven-day weeks going to zero is “the worst thing in the world,” but “it’s a lot easier to build a company the second time around” — the first firing, the interview process, the pipeline, all the things that are easy to screw up once.

2. Cows to sepsis to COVID — one technology, three TAM lessons

• The through-line is DNA testing, and the search rule is market-first: “if you’re looking to do better DNA testing, you’re just looking for markets where people care more about that.” Cattle genomics (TL Biolabs) started when a farmer he’d told no mailed samples “with a check attached to the front.”
• The Series A problem was arithmetic; he says the bio fund hadn’t done the TAM calculation at seed: ~100 million cows in the US at $15–20 a test is a $1.5 billion total market even if you test every animal every year — “not enough to do a Series A off of.” His verdict on the $1.65M seed: “I don’t think they did the TAM calculation.”
• Sepsis was the next TAM upgrade, and Turner’s framing of the clinical stakes is the load-bearing number: what kills you isn’t the bacteria but your own immune cascade, and “every hour that you don’t treat somebody is about a 12% increase in mortality.”

3. “It doesn’t matter if you have a better test if no one ever runs it”

• Curative’s original pitch was sepsis again, from a different angle. Turner’s read of the graveyard — likely Roche a couple hundred million, likely Siemens a hundred million — is that the tests work in academic medical centers and fail in the bigger community-hospital trials. Not because the science breaks. Because at 2am on a Saturday in an understaffed ER, nobody suspects sepsis early enough to order it.
• Hence the plan: a “mini hospital in a hospital” that takes the sepsis patient out of the emergency room on a fixed fee — “so no matter what happens we’re on the hook” — mostly by getting doctors to follow the instructions at scale.
• The first signal of what was coming was a put-on-hold pilot. Their enthusiastic Wisconsin clinician went silent and had his assistant say everything was on hold for three months. “That was the first inkling for me of like, oh crap, this is going to be a big thing.”
• Justin Mateen was first money in at Curative — $125,000 at a $3 million valuation — and, per Turner, “he didn’t even know what it was.”

4. The incumbents’ core competence was the disqualifier

• Asked what he saw that the big labs didn’t: Quest and LabCorp are “ultra efficient machines” optimized to win 1% of margin on a process step, “but if you’re asking them to 10x capacity, that’s literally the opposite of what they’re built for.” The mindset isn’t there and the supply chain broke anyway.
• So: “start from scratch, throw all of that away” — the “orthogonal supply chain,” which Harry immediately clocks as consultant-speak (“sound like a McKinsey consultant specializing in innovation”), and which Turner cheerfully concedes was “a good fancy word… helpful from a sales standpoint.”
• What it actually meant: everyone bidding for the same consumables doesn’t net-add a single test, “it just makes us all squabble over it.” So they bought swabs from electronic-testing vendors and sterilized them, and swapped magnetic beads for filter plates — “basically just glass and plastic” — because beads “all come from basically two factories in China.” Turner is explicit this only works inside a pandemic.
• Curative won the Florida state nursing-home contract — every employee at every nursing home, once a week, for three months, “in the hundreds of millions of dollars” — after the rest of the field bid no. “Everybody said, ‘No, that’s too crazy… this is impossible.’ And we bid… And we delivered it.”

5. $5 billion of revenue on a business that lost money between waves

• Growth mechanics, as told: Laura Deming drove him to LA with a car full of PCR machines and tweeted that they had capacity; the deputy mayor of LA “slid into her DMs.” Because the city paid after delivery, they invoiced daily and sent someone to City Hall every morning to pick up a check — net-1, not net-30 — because that cash bought the supplies to scale.
• Hiring at that speed meant people lined up socially distanced down the street in the parking lot with “somebody sit there with a clipboard” running five-minute interview slots. Seven employees to 7,000 in nine months; ~$5 billion over three years; the largest likely non-LabCorp testing company.
• The margin structure is the part investors should keep. Peak surge — the Dodger Stadium site running seven lanes, 7am to 7pm, seven days a week, ~10,000 people a day — is very profitable; the dips leave you at 20–30% utilization with identical fixed cost. “We would lose money on every test during the dips.” Reimbursement was priced high on purpose to make someone overbuild the peak, “because if you don’t build that peak capacity then when you have a surge it all goes horribly wrong.”
• On vaccines, two separate hedges worth keeping exactly as hedged. On efficacy: “it was not clear at the beginning,” benefit fell once everyone had COVID a few times and variants weakened, but “in December of 2020 there was benefit for a lot of people getting the vaccine.” On the business: 2.5 million vaccinations, “we lost money on every single dose. It cost more to administer them than we were getting paid.” Why do it? “Giving back. A lot of our partners wanted it.”

6. Everything in US healthcare routes back to the payer

• The search for act two ruled out its own industry on TAM: even displacing all of LabCorp and Quest is “about 30 billion of market cap,” and “coming out of what we did with COVID I wanted to build a much bigger company than that.”
• Buying hospitals — one in Florida, one in Texas — was ruled out on structure, not ambition: “the payer mix is too broken up” — half government, the rest fragmented plans that “change their mind every 5 minutes” — so your ability to change anything from the provider side is limited. “The payer is the one that drives behavior in the US healthcare system. If you are providing the dollars, people will go where the dollars are.”
• His magic-wand fix is anti-consolidation: healthcare is “quite an efficient system as a market when the counterparties are small.” The specific: a primary care doctor affiliated with a hospital system gets paid on average double an independent one for the same service with the same credentials, because the system bundles its beds and surgery centers into the negotiation. With four large payers against mega-systems, “they just reach these loggerheads where nothing gets done and everybody’s overpaying.”
• Categorical, and he doesn’t hedge it: is the US the best place to be treated if you’re sick? “Yes, definitely… by far the most cutting edge techniques and facilities and drugs than the rest of the world.”

7. “I wish that I knew AI was coming”

• The stated regret about designing the insurance business in 2022: “we had no idea that this wave of AI and LLM was coming… in the last like 18 months, how we do pretty much everything is now a completely different workflow.” His framing of why insurance is so exposed: “all health insurance does is like moving bits around” — a plastic card, a database, and a marketplace between providers, employers and members.
• The cleanest before/after is credentialing — verifying licenses, transcripts and malpractice records for network doctors. Five or six people, two to three months, ~$50. Now an in-house agent running on Claude does it end to end: 12 hours, about 20 cents.
• The generalizable trick, and it’s a good one: models “are not very good at parsing files. But they’re incredibly good at codegen.” So for broker submissions arriving as arbitrary PDFs and spreadsheets, the agent writes a single-use Python script, loops until it converts the file to the standard format, and then the script is thrown away. Result: “Now send us whatever you’ve got, whatever format. It can be scribbles on a napkin, it doesn’t matter.”
• Harry pushes the VC assumption that data cleanliness is the enterprise blocker. Turner says no — “the models are so good at cleaning up the data if you give them the right context.” Evidence: a Looker-to-Snowflake migration that “would have taken like probably like a year and a whole bunch of engineers” ran as an agentic loop converting dashboards one at a time, done by one or two people in a couple of months.

8. Gwen: from 100 contracts a week to 100 a day

• The structural moat Curative is attacking: about 1.2 million providers in the US, 60–70,000 contracts, built by Blue Cross over a hundred years and likely UnitedHealth over fifty. “This has been one of the biggest pieces of staying power of the big health plan businesses.” Curative’s 45-person contracting team did 2,300 contracts last year.
• Gwen does the whole workflow from a lead: Googles the practice, reads the price-transparency files to see what other payers pay them, finds the email on ZoomInfo, writes customized outreach, negotiates rates over multiple rounds, redlines the language (again via Python, because “these models are terrible at editing Word documents”), and signs with Turner’s signature on DocuSign. In about eight weeks the agent alone did 3,500 contracts.
• Two economics: $1,500–2,000 per contract with people, about $70 with Gwen — which is why a 2x or even 5x Anthropic price rise wouldn’t change the calculus. “If I say yes, I don’t want our Anthropic rep to double our pricing,” Turner jokes, then answers anyway: “It would be fine.”
• The counterintuitive edge isn’t cost, it’s shamelessness: Gwen sends ~15,000 customized emails a day and “a lot of providers will get them on the ninth email. There’s no way that a human is going to email them nine times.” The team didn’t go to zero — it was moved onto the big hospital systems that want a phone call and a dinner. Harry’s contribution: he got likely Marc Benioff on the show by emailing him 53 times. “I’m basically an AI model. I lost my personality.”

9. The SaaS cut, the MLR trap, and why the incumbents can’t follow

• Harry says he buys the theory that SaaS is dead: “Because I see the number of contracts we’re canceling.” Salesforce: $600,000 a year to zero, replaced in two months by a vibe-coded internal CRM that agents run inside. SaaS spend is being cut ~80% this year, managed off an internal slide of renewal dates and who has to make the call.
• Harry’s pushback is the right one — is it worth the engineering hours to build and maintain it? Turner concedes maintenance is “definitely one of the most challenging pieces,” but points at the fully-loaded alternative: a full-time Salesforce administrator plus other tools “taking multiple FTEs to maintain,” versus “one great engineer” and any custom feature you want. What survives is infrastructural: Sentry, and Slack because of accumulated integrations — “if they put the prices up too much, then eventually it’ll make sense to replace that.”
• Why the big payers can’t just copy this: it’s people-ops, not technology. “If you’ve built a company of 100,000 people and in order to get this margin improvement 50,000 of them have to be laid off, somebody’s fiefdom just got a lot smaller. And so they will do it slowly over 10 years.”
• The regulatory perversity he wants on the record: 85% of premium must go out the door as care by law, so profit is capped at 15% and “the only way to increase profits is to increase total spending.” Harry asks why he would encourage people to go to the gym; Harry calls capping insurance company profits “a great PR thing to say,” but says it’s BS.
• Headcount call: 650 today, “quite a bit smaller” near-term — Harry floats 400, Turner: “somewhere around there” — bigger in five years. Care navigators are the exception and scale linearly with membership, because “there has to be a person somewhere in the loop.”

10. Subcritical: nuclear is a regulatory problem wearing an engineering costume

• The company he co-founded with his wife (they exchanged genome VCF files after the first date to check compatibility — “I do carry the ginger gene”) exists because everyone else was solving the wrong problem: “it’s not a science or engineering problem… we have built safe nuclear reactors since the 60s.” The blocker is a regulatory environment shaped by the anti-nuclear push in the 1980s, including a demand to guarantee nothing goes wrong in a once-in-a-million-year event.
• The design answer is likely Carlo Rubbia’s late-’80s “energy amplifier,” and Turner explains it as a knife-edge you simply step off: conventional reactors must balance at exactly 1.0 criticality; Subcritical runs at 0.97, so the reaction always fizzles, and a particle accelerator supplies the missing neutrons. Turn the accelerator off and output stops. “Even if you put in 10 times as much power from the accelerator, it will never run away.”
• Scale and financing, stated plainly: about $1 billion of construction cost per 300MW facility, funded as infrastructure with equity and debt rather than venture equity. The only such system under active construction is in China, “based on a US design from the 2010s that the US stopped working on after Fukushima.”
• Asked which company gets bigger, he picks Subcritical — while conceding Curative has the larger market opportunity, since US employers spend $1.5 trillion a year on healthcare and addressable nuclear energy is “a similar order of magnitude.” Harry: “They’re both yield optimization.” Turner: “I figured out the TAM thing.”