About
I'm Will, a software engineer based in London and the CTO at Flowstate. We're building…
Articles by Will
-
Portability should be a factor when deciding on cloud solutions
Portability should be a factor when deciding on cloud solutions
Businesses are adapting cloud—it's a fact. We're well past the point of deciding whether or not the cloud is a fad, and…
1 Comment
Activity
2K followers
-
Will Hackett reposted thisWill Hackett reposted thisAI Labor Is 25% of Human Labor cost, tracking to 100% at SemiAnalysis. This is a whole new workforce - that scales in cost at speed - fundamentally changing cost structures. If you’re growing in revenue, that’s no sweat. But if you’re not? How do you make decisions about where to invest? (in flowstate of course)
-
Will Hackett posted thisClaude's personal plans are substantially better value than Enterprise or Teams, which quickly tip into raw API spend. A recent teardown of Claude's backend limits (via some cleverly reverse-engineered floats) shows just how much bang for buck you get from individual subscriptions vs the API: • Pro ($20/mo): ~$163 in equivalent API value (~8x) • Max 5× ($100/mo): ~$1,354 (~13.5x) • Max 20× ($200/mo): ~$2,708 (~13.5x) Cache reads are 100% free on personal plans, while the API charges 10% per read. If your devs are running agentic loops (like Claude Code) against a warm cache, that multiplier skyrockets to an insane 36x. Per-seat usage limits on Team and Enterprise are actually substantially lower than a personal Max plan. You're very suddenly bleeding cash on raw API usage to keep them moving. The optimal playbook for engineering teams right now: 1. Issue virtual cards so your engineers can subscribe to their personal AI tool/plan of choice 2. Use Flowstate to track where the spend is going and set policy on how it's used It requires a slight shift in how you manage tooling, but the hard-dollar savings for heavy-usage teams are impossible to ignore. $3m+ p.a. for a 100 engineer team.
-
Will Hackett shared thisLondon. New York. Flowstate is hiring. If you want to define a category and ship product that matters, we're your team. https://lnkd.in/eaMfuNk9
-
Will Hackett shared thisWe're hiring a Product Engineer at Flowstate. The best product engineers we know have two things in common: they obsess over customers, and they're just as comfortable debugging a slow query as they are debating a colour palette. If you can hold both of those things at once (and you want real ownership over what gets built) we'd love to talk. London, Moorgate. In office. Free lunch. Serious problems. Details in the console 👉 https://www.flowstate.inc/
-
Will Hackett shared this"Wait... how the f*** did we spend $990k on tokens this month?" If you're an engineering leader, you've probably had a similar panic attack. We got tired of flying blind. So we built a way in Flowstate to attribute every single token directly to specific items in your backlog. Now you actually know what your team is building with all that compute.
-
Will Hackett shared thisFantastic evening at the Databricks 'AI in Production' event last night. There were a lot of great insights, but Piero Sierra absolutely stole the show for me. He shared three perspectives that really resonated with how we need to be building right now: AI Chat is a UI failure state: If your user has to stop and ask a chatbot how to do something, your core product interface has failed them. AI should enhance the experience, not act as a band-aid for clunky design. Theory of Constraints. Applying this to modern AI product development is a masterclass in figuring out where the actual bottlenecks are, rather than just chasing hype. Stop calling it "vibe coding." It’s just "coding" now. Traditional, non-vibe coding is the old thing. This is just how we work today. It's incredibly refreshing to hear a product leader at Skyscanner's scale talk about AI with such pragmatism and focus on the actual user experience.
-
Will Hackett shared thisI need to issue a public correction. I WAS WRONG. My "600 photocopiers" metaphor was too small. Sequoia Capital just proved the ratio of AI software to AI services is 1:6. It’s 3,600 photocopiers. 🖨️🖨️🖨️ Because we are no longer buying software. We are buying a parrot that read the internet. We are trying to get an itemised receipt from a bird. Paying premium rates to a sentient blender. Amortising a hallucination. You aren't just replacing Jim from Accounting. You are replacing Jim with an algorithm that has no concept of what money actually is, and legally has to be classified as operating expenditure. Link to the full rant is in the comments. May God have mercy on your Q3 margins. 👇
-
Will Hackett shared thisI want 600 photocopiers. 🖨️ 🖨️ 🖨️ 🖨️ That's essentially what some companies are doing with AI spend right now, and I feel like we aren't gossiping about it enough. "We need to move faster." "AI can accelerate everything." "Let's spin up 52 projects this year." Cool. You've just ordered 600 photocopiers. You don't have 600 offices. You don't have 600 people who need to print things. You don't even know if the 600 things you're about to print are things anyone wants to read. But the photocopiers are on order and the invoices are coming. Reality's coming! Team A is burning $250K+ a month on AI tooling. Shipping constantly. 52 projects a year. The energy is incredible. The Slack channels are full of memes. The AI bill is absolutely eye-watering. Revenue from all this output? Unclear. Possibly great. Possibly... just... vibes... Team B is spending a quarter of that. Shipping 12 things a year. Each one is trackable. Each one connects to an outcome. The AI bill is boring. The CFO can actually explain it to the shareholders. Everyone wants to be Team A. I'd argue Team B is more likely to win. Not because shipping fast is bad. It's not. AI-powered delivery velocity is genuinely transformative. But velocity without measurement is just expensive movement. And here's the bit that really hurts: if you can't attribute that AI spend (to CapEx, to R&D, to capitalised software development) it lands on your P&L as a blob of operating expense. Your margin nosedives. The board sees cost growth they can't explain. You miss R&D tax credits worth 6–10% of qualified spend. You're outpacing your own revenue with invisible cost. Gartner says $2.52 trillion in global AI spend this year. Only 14% of CFOs report clear ROI from it. Just over half say they can even track it. The question isn't "how much should we spend on AI?" It's "how much of our AI spend can we actually connect to revenue?" You don't know the answer until you've tried both models. And you can't optimise either one without the data. We built the first AI modelling tool that can handle this attribution — connecting engineering delivery to financial outcomes. So your CFO doesn't have to guess whether the 600 photocopiers were worth it. They probably weren't. But now you'll know.
-
Will Hackett shared thisToday, a technology leader at an enterprise told me he's under pressure to show savings from AI usage. They won't be the last. Your engineering team's AI spend is going up... a lot. Most companies are expensing all of it as OpEx. That's probably wrong. (I'm no accountant, but I am related to a very good one.) If your engineers are using Claude Code, Copilot or ChatGPT to build and ship software, then the cost of those tools can be capitalised against specific projects. The same way you capitalise engineering salaries. Whether it's UK's FRS 102 s18 or the US ASC 350-40... the principle is the same: development costs that create future economic value get capitalised. The problem is nobody can attribute the spend. Your Anthropic invoice doesn't tell you which team used it, on which project, for what purpose. So finance lumps it into general "OpEx" and moves on. That matters more than you think. Engineering teams are now spending $500–$3,000+ per developer per year on AI tools and that's before agentic workflows really scale. When AI spend bloats 3–5% of your engineering budget, it stops being a rounding error and starts being a line item your CFO and board want to understand. The companies that can break this down by project don't just get cleaner P&L... they get more room to invest. A CTO who can say "we spent $80k on AI tooling last quarter, $52k of which was capitalised against these three product initiatives" is having a fundamentally different conversation with the board than one who says "yeah, AI costs went up." That's what we're building at Flowstate.
-
Will Hackett reacted on thisWill Hackett reacted on thissaw someone vibe coding at a cafe no multi-agent setup no 3-hour extended thinking loops no switching between Codex and Claude Code just prompting, and staring at the screen, waiting for the response like a psychopath
-
Will Hackett liked thisWill Hackett liked thisAI Labor Is 25% of Human Labor cost, tracking to 100% at SemiAnalysis. This is a whole new workforce - that scales in cost at speed - fundamentally changing cost structures. If you’re growing in revenue, that’s no sweat. But if you’re not? How do you make decisions about where to invest? (in flowstate of course)
-
Will Hackett liked thisWill Hackett liked thisWe're hiring engineers to keep up with the pace of demand. AI is the fastest-growing and least-governed cost in the enterprise. We're helping companies maximize (and control) their AI investment. Come join us!
-
Will Hackett liked thisWill Hackett liked thisAnother incredible brand moved over to our Status Page product today. In the hype-y world of AI, it's easy to forget the importance of basics like communicating with your customers in an outage. But in a world where things go wrong all the time, it remains a pretty key differentiator in the overall quality of customer experience. Welcome to the club Plaid. You're in great company... OpenAI: status.openai.com Lovable: status.lovable.dev Linear: linearstatus.com Intercom/Fin: www.intercomstatus.com Etsy: status.etsy.com
-
Will Hackett reacted on thisWill Hackett reacted on thisanyone else getting this issue with their AI assistants??
Experience
Volunteer Experience
Other similar profiles
Explore more posts
-
Edward Tandia
Fundraising doesn’t fail because of a bad pitch. It fails because of misalignment. In our latest episode, we spoke with Tushar Kansal, Founder & CEO of Kansaltancy, about what really shapes strong founder–investor relationships over time. We discussed: -Why “being investor-ready” is more than financials -How founders can avoid raising the wrong kind of capital -What investors pay attention to once the deal is done -How early clarity prevents long-term friction This wasn’t a conversation about fundraising hacks. It was about alignment, expectations, and building partnerships that last beyond the raise. If you’re a founder navigating capital decisions or an investor backing early-stage teams this episode is worth your time. 👉https://lnkd.in/ebVmPBt8
4 Comments -
Eric Faw
This was always the natural next step for us. AI only becomes useful in fund accounting when it’s grounded in deterministic definitions - capital, allocations, ownership, FX, hierarchy. Without that, outputs aren’t something teams can rely on. Entrilia’s been built around that semantic foundation from day one, which is what now allows agentic workflows to run inside real accounting processes. If you’re a GP or fund administrator and want to see how this works in practice, we’re walking through it here 👇
2 Comments -
Dirk Strauss
This made me chuckle - but there’s definitely some truth here. It's amazing how everything, from dishwashers to wealth advisors, is now 'AI-enabled'. Having an 'AI inside' label on the box seems essential for attracting funding, even though there’s strong evidence emerging that many businesses (95% based on a recent MIT study) are still struggling to achieve meaningful ROI on AI spend. That said, at Chisl, we’ve seen firsthand how revolutionary the tech can be when applied correctly. I am definitely a buyer of the hype!
1 Comment -
Johan Surani
It’s only 13 days into the year but I think I just watched one of the best interviews on venture investing for the year. Harry Stebbings interviews Alastair (Alex) Rampell who leads a16z’s $1.7B apps fund and they talk about a range of investing related topics. A great summary is also included in Harry Stebbings’s post below.
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More