@BessemerVP
For the entrepreneurs who want to build revolutions of their own.
1/ We spent H1 2026 talking to 20+ CTOs about agentic AI adoption in engineering teams. On average, organizations spend $1-2K per engineer per month on tokens (many spend less!). But the average hides a wide internal distribution: a few heavy parallel-agent users at $8-10K/month, many lighter users near zero. Hint: @liranesh says if you’re spending <$1K/month, you’re not there yet. Dive into the token trends to see the real story. 🧵
2/ Teams often have super-engineers producing extraordinary output—or inefficient token-maxxing. The only way to separate them is through quality-adjusted throughput and cost per accepted result. But high spending doesn't tell you which.
Code generation revved up your devs by 10x. But in every product review, PM handoff, and standup, you’re still hitting “red lights.” You bought the Ferrari, and now it’s time to redesign the roads and choose new routes. 𝐆𝐞𝐭 𝐭𝐡𝐞 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐧𝐠 𝐦𝐨𝐝𝐞𝐥 𝐭𝐨 𝐚𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐞 𝐚𝐠𝐞𝐧𝐭𝐢𝐜 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 → t.co/akAwYDUNVW
Agentic AI widens the security surface, with risks coming in four directions at once. → Risk in what agents write → Risk in what agents do → Risk from attackers’ agents → Risk in the model and provider The 10 Commandments for Agentic Security are essential for any team running agents. Learn more with 𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐰𝐚𝐤𝐞𝐧𝐢𝐧𝐠 → t.co/akAwYDUNVW
A new episode of @hoh_pod is live! @stephenkraus talks with Dr. Michael Gao & Dr. @joshgeleris, co-founders of @DxSmarter about 18 months with no customers, a pivot that made no sense on paper, & the path to a billion+ acquisition. Listen: bessemervp.team/4xksoAs
"𝟗𝟎% 𝐀𝐈-𝐰𝐫𝐢𝐭𝐭𝐞𝐧" ≠ 𝐚𝐠𝐞𝐧𝐭𝐢𝐜 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 Without autonomous-agent infrastructure (multiple parallel agents per engineer) you're still capped by human attention. That's an IDE plugin with a good marketing line. AI-coding tools alone are not what transforms developer teams—it’s about the operating model along with it. Infrastructure enables the speed, but the people and org structure decide whether it actually compounds. Learn more with 𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐰𝐚𝐤𝐞𝐧𝐢𝐧𝐠 → t.co/akAwYDUNVW
𝐇𝐚𝐬 𝐲𝐨𝐮𝐫 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠 𝐭𝐞𝐚𝐦 𝐜𝐫𝐨𝐬𝐬𝐞𝐝 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐭𝐢𝐜 𝐭𝐡𝐫𝐞𝐬𝐡𝐨𝐥𝐝? Here are a 5 signs: → $1K+/month per engineer in tokens → Top 1% of engineers ship 46× the median → Org charts widen: ~15-25 reports per manager → The PM-to-engineer ratio breaks into one extreme: 1:10 where eng absorbs the PM role, or 1:1 where every eng needs a PM → Four chiefs run the show: Architect, Product, Designer, Security Read this before your next eng offsite ⭐️ 𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐰𝐚𝐤𝐞𝐧𝐢𝐧𝐠 → t.co/akAwYDUNVW
📣New @hoh_pod: @stephenkraus & @halletecco talk with @StanfordMed's Dean, Dr. Lloyd Minor on how AI is reshaping medical education, why tumor boards no longer start with humans, and what's still blocking precision medicine. 🎧Listen: bit.ly/4x4mkM7
𝐓𝐡𝐞 𝐜𝐨𝐝𝐞 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐜𝐨𝐧𝐮𝐧𝐝𝐫𝐮𝐦: engineering got close to 10× faster, but the organization shipped less than 50% more. This gap is the reason @liranesh and @AdamRFisher wrote The Agentic Awakening. ⭐ The leap to agentic engineering requires transformation at the operating model. 𝐆𝐞𝐭 𝐭𝐡𝐞 𝐩𝐥𝐚𝐲𝐛𝐨𝐨𝐤 → t.co/akAwYDUNVW
We talked to CTOs and engineering leaders at 20+ frontier companies, including @tryramp, @Lemonade_Inc, @wonderful_ai, @drivenets, and more, all pioneering a new way to set the pace of innovation and shipping. The bottom line? The fastest movers don't just build AI infrastructure. They rebuild their entire organization around it. Read 𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐰𝐚𝐤𝐞𝐧𝐢𝐧𝐠 ⭐️ t.co/akAwYDUNVW
Today we’re launching 𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐰𝐚𝐤𝐞𝐧𝐢𝐧𝐠, the deep-dive research and playbook on agentic development. ⭐ 20+ AI-pilled engineering teams → 63-pages of insights and frameworks to guide a company’s transformation. Developer teams want to know: "How are we measuring up? 🧵 t.co/akAwYDVlLu
2/ We talked to CTOs and engineering leaders at 20+ frontier companies, including @tryramp, @Lemonade_Inc, Wonderful, @drivenets, and more, all pioneering a new way to set the pace of innovation and ship. The bottom line? The fastest movers don't just build AI infrastructure. They rebuild their entire organization around it. Here’s how. ⤵️
A little summer @hoh_pod replay ICYMI the first time around: Listen as @stephenkraus and @mtesquivel sit down with @ouraring CEO @tomeghale to discuss clinical accuracy, overnight tracking, and Oura's fast lane into real healthcare. bit.ly/4fSXFmk
@recallai @strella_io @GraphAIOfficial @ada_cx @davidgu 6/ At @strella_io they use seat-based pricing over per-interview pricing to maximize adoption, not revenue per use. Why? Because AI costs pale in comparison to paying research participants. When delivery economics are favorable, founders have flexibility to optimize for growth.
7/ At @GraphAIOfficial they price the labor, not the software. Graph AI benchmarks pricing against the combined cost of software + human review, not competitor software tools. They automate 75% of a workflow previously billed by headcount. The unit of pricing follows the unit of value, not the technology.
4/ Hard ROI commands premium pricing. Soft ROI compresses it. A copilot that "surfaces the right answer" leaves the outcome in the customer's hands. Value is real, but soft, leaving customers to ask "would this have happened anyway?" With an agentic product that closes the loop end-to-end that question disappears and pricing power follows.
3/ Three variables determine your pricing power: 1️⃣Customer value: How much value are you creating for your customer? 2️⃣Fungibility: How defensible is your product? 3️⃣Delivery economics: What does it actually cost to serve customers? Get this mix right and everything else follows.
2/ Over the past year, AI companies have experimented with usage, workflow, outcome, and hybrid pricing models. Increasingly, businesses gain the most revenue momentum by pricing on outcomes, but the cost-benefit tradeoff depends entirely on the ideal customers they serve.
1/ How should you price your AI product? We took a look inside four companies—@recallai, @strella_io, @GraphAIOfficial, and @ada_cx—to understand how they each approached pricing and what made their strategies durable at renewal time. Here's what we learned. 🧵
@recallai @strella_io @GraphAIOfficial @ada_cx 2/ Over the past year, AI companies have experimented with usage, workflow, outcome, and hybrid pricing models. Increasingly, businesses gain the most revenue momentum by pricing on outcomes, but the cost-benefit tradeoff depends entirely on the ideal customers they serve.
𝐍𝐞𝐰 𝐟𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤 📈 Bessemer’s AI-Native Services evaluation framework. Our investment team shares the 11-criterion score card they use to measure market attractiveness. Starting with → 𝐌𝐚𝐫𝐤𝐞𝐭 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞 𝐚𝐧𝐝 𝐝𝐞𝐦𝐚𝐧𝐝🧵 1. Vendor fragmentation 2. Incumbent response capacity 3. Criticality and failure cost 4. Demand elasticity under AI pricing 5. Supply side labor scarcity
@cbirn @eric_kaplan_nyc @maha_a_malik_ @Libbiefrost 𝐃𝐞𝐥𝐢𝐯𝐞𝐫𝐲 𝐞𝐜𝐨𝐧𝐨𝐦𝐢𝐜𝐬 6. Labor share of COGS & automation ceiling 7. Outcome verifiability and error tolerance 8. Pricing model & surplus capture
Eight years ago, @maintainx started at a kitchen table in a small San Francisco apartment. 🌉 Despite the naysayers, co-founders @CTurlica, @dozoisch, @theNickHaase, and Mathieu Marengère-Gosselin believed they could build a better platform for industrial teams and frontline workers.
In 2021, @bdeeter led Bessemer's investment in @maintainx, fully backing and believing in their vision to turn asset and work data into real-time intelligence for industrial operations.
Eight years ago, @maintainx started at a kitchen table in a small San Francisco apartment. 🌉 Despite the naysayers, co-founders @CTurlica, @dozoisch, @theNickHaase, and Mathieu Marengère-Gosselin believed they could build a better platform for industrial teams and frontline workers.
In 2021, @bdeeter led Bessemer's investment in @maintainx, fully backing and believing in their vision to turn asset and work data into real-time intelligence for industrial operations.
Two Bessemer companies, one vision. 🛠️ Seven years ago @davidcowan bet robots would learn to see and understand the physical world. Today that vision came to life: @procoretech announced it has entered into a definitive agreement to acquire @DroneDeploy, uniting construction's system of record with its system of perception. Congrats to @mikewinn, @jonomillin, @nickponline and the whole DroneDeploy team!
Thank you to all involved in the journey of these companies coming together! CC: @davidcowan, @brianfeinstein, @chrisxwan, @tesshatch, and many more across our Bessemer team and alumni.
1/ How do you pick the right $1B+ problem to solve? There are typically three ways founders fumble at the early stage: 1️⃣They pick the wrong problem to solve 2️⃣They don’t understand their ideal customer profile (ICP) well enough 3️⃣ Some combination of the above The path to a billion-dollar business starts with getting these early decisions right. 🧵
2/ Choosing the right market When it comes to choosing a market, the tension isn't between a big TAM and a focused entry point—it's how you thread the needle between them. @AdamRFisher encourages founders to target a growing subset of a large market rather than prematurely universalizing the product.
Biology-native data infrastructure is what will separate the AI-driven biotechs that scale from the ones that stall. As models improve and compute gets cheaper, that infrastructure (not the model alone) is where the real edge comes from. Here's how we think about the data, the workflows, and the lab automation that will define the next generation of leading biotechs, and the companies already building in the space. 🧵
2/The Cambrian explosion of AI models for biology has already happened: from single-digit model releases in 2015 to 380+ new biology AI models in 2025.
Corporate America is predicated on trust, data security, and collaboration with their vendors, says Byron Deeter. “[Fortune 1000 companies] are not going to chase fractions of pennies per token and embrace Chinese technology for their most critical data assets, market needs and product needs.” — @bdeeter Learn more from @CNBC: t.co/dVRtRpGUex
On Bloomberg, @spdholakia joins @EdLudlow to discuss: → Why competition is good for AI → Where the frontier model makers still have an edge → How AI Giants are reaching the $1B ARR milestone 2x-3x faster than the SaaS cohort. Watch it all → bloomberg.com/news/videos/20…
1/ Last week @LindseyLi_ hosted @mattyp and @DevRelChap for a conversation about all things DevRel in an agent-first environment. Here are the main takeaways. 🧵
2/ User empathy is the foundation, and it scales to agents. Build content around the exact problem your user is solving, not the category your product sits in. @HashiCorp's top content was never about "secrets management." It was "how to rotate database credentials."
𝐅𝐢𝐫𝐞𝐰𝐨𝐫𝐤𝐬 𝐠𝐫𝐞𝐰 𝐟𝐫𝐨𝐦 $𝟏𝟎𝟎𝐌 𝐀𝐑𝐑 𝐭𝐨 $𝟏𝐁 𝐀𝐑𝐑 𝐢𝐧 𝟏𝟔 𝐦𝐨𝐧𝐭𝐡𝐬! @AnthropicAI set the revenue trajectory for the next generation of AI Giants. And @FireworksAI_HQ demonstrates the path to $1B ARR in 4 years, isn’t a fluke, but a new possibility. 𝐅𝐢𝐫𝐞𝐰𝐨𝐫𝐤𝐬 (𝐉𝐚𝐧 𝟐𝟎𝟐𝟐) 𝐯𝐬. 𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 (𝐉𝐚𝐧 𝟐𝟎𝟐𝟏) — 𝐅𝐨𝐮𝐧𝐝𝐢𝐧𝐠 → $𝟏𝟎𝟎𝐌 𝐀𝐑𝐑: Both took ~3 years — $𝟏𝟎𝟎𝐌 → $𝟏𝐁 𝐀𝐑𝐑: Fireworks (16 months) vs. Anthropic (11-12 months) — 𝐅𝐨𝐮𝐧𝐝𝐢𝐧𝐠 → $𝟏𝐁 𝐀𝐑𝐑: Fireworks (~4 years) vs. Anthropic (~4 years)
1/ Last week @LindseyLi_ hosted @mattyp and @DevRelChap for a conversation about all things DevRel in an agent-first environment. Here are the main takeaways. 🧵
2/ User empathy is the foundation, and it scales to agents. Build content around the exact problem your user is solving, not the category your product sits in. @HashiCorp's top content was never about "secrets management." It was "how to rotate database credentials."
2️⃣Full-stack builds deeper customer insight, but consider partnerships as well Owning hardware, deployment, and the customer relationship gives you the tightest feedback loop. But as the ecosystem matures, strong integration partners are becoming a real strategic option.
3️⃣Don't lock into a vertical or a stack too early Hardware isn't reliable enough yet to cleanly split "foundation" from "brains." Until it is, owning more of the stack and iterating on everything together beats specializing too soon.
