Hi Reader

This edition contains several variations on the same story: “in theory, this should work”. In practice, as is often the case in business (and in life generally) it turns out to be slower, messier or more revealing than expected. But as a successful founder once said to me: if launching a business was easy, everyone would be doing it.

Enjoy!

Is AI actually being used at work?

After years of spending, pilots and promises, new data suggests something more awkward is happening inside organisations: declared AI use at work may be slipping rather than accelerating. US Census Bureau surveys, analysed by The Economist, show a growing share of firms reporting that they are not using AI to produce goods or services, including large employers that were expected to lead adoption. The signal here isn’t that AI has failed, but that deployment is proving hard. Fatigue, poor integration and unclear ownership appear to be catching up with the hype, raising an uncomfortable question for leaders investing billions in infrastructure: if usage stalls, where does the return actually come from? Futurism has the story.​

Voice search is becoming a buying behaviour

Voice assistants are no longer confined to weather checks and timers. Research shows they’re increasingly embedded in everyday search and shopping, particularly among younger, higher-income users who value speed and low friction. Around a third of consumers now use a voice assistant weekly, with significant overlap between voice users and people actively experimenting with AI tools such as ChatGPT. What makes this shift commercially relevant is intent: voice users are more likely to have purchased online recently and to use voice for actions, not just information. For many users, voice isn’t about convenience at all. It’s an accessibility tool – and that shifts it from novelty to infrastructure. For brands, the question is whether their search, content and checkout journeys are designed to be spoken rather than typed. Find out more here.

When selling the company isn’t the end of the story

I came across an interesting HubSpot piece which is one of the more honest accounts of what acquisition feels like from the founder’s side. George Davis from Frame AI describes the trade that sits behind most deals – giving up independence in return for scale – and what makes it worth reading is that it doesn’t position acquisition as a victory lap. It treats it as a compression of possibilities into a single path and asks whether the impact you want is bigger than the title you’re giving up. For anyone building with one eye on an exit – or wondering what comes after – it’s a grounded, human account of choosing scale over control. Read it on HubSpot here.

Checkout moves into the chat window

Microsoft is testing what happens when buying no longer requires leaving the interface where a decision is made. Its new Copilot checkout lets users search, compare and complete a purchase inside the chat itself, with payments handled via partners such as Stripe and Shopify. On the surface, this looks like a convenience feature. More interesting is what it signals: search, recommendation and transaction are collapsing into a single step, owned by whoever controls the interface. For merchants, Microsoft is keen to stress they remain the merchant of record and keep customer data. For everyone else, it’s another sign that AI assistants are positioning themselves between brands and buyers at the moment of intent. This rolls out in the US first, but the direction of travel is clear. Find out more here.

Reed Hastings on the Netflix culture code

Netflix co-founder Reed Hastings is at his most interesting when he strips away the mythology around the brand and talks plainly about trade-offs. In a YouTube conversation, he unpacks why “talent density” only works if leaders accept turnover, how organisations drift into false consensus and what the Qwikster decision taught him about the cost of suppressed dissent. He’s particularly clear on how boards function in practice – not as a source of day-to-day wisdom, but as an insurance mechanism when leadership decisions fail. It’s a long interview, but a grounded one, with plenty of insights. Watch it here.

This might even convince me to start running again

What a brilliant little January idea from pizza chain Franco Manca: map your run so it looks like a pizza (or a pizza slice) and they’ll give you £1 off a pizza for every kilometre you’ve run. It taps into January’s get-fit resolve, keeps sales going during a quieter time and is instantly understandable by customers. And it’s fun! Check it out here.

Why AI agents struggle with the small stuff

A detailed piece from Amazon’s AGI Lab explains what sits between “book my holiday” and a working AI agent. Before an agent can plan anything, it has to master hundreds of low-level, failure-prone actions – scrolling, clicking, recovering from broken forms and legacy systems – and do them correctly every time. The article walks through how Amazon trains agents on these tiny tasks using reinforcement-learning “gyms”, and why reliability, not intelligence, is the real bottleneck. It’s technical in places, but it’s a great explanation of why agentic AI is harder than the demos suggest. Read it here.

What Baileys’ accidental invention teaches about product discovery

The iconic cream liqueur now known as Baileys began not in a lab or boardroom, but from a vague brief and a few kitchen experiments. In a first-person account for The Irish Times, the co-creator describes how a throw-together mix of whiskey, cream and chocolate went from uncertain prototype to global export – even when early focus groups were lukewarm at best. For anyone wrestling with market ambiguity, iterative prototyping and the difference between early feedback and eventual adoption, it’s shows that rigid testing isn’t the only path to products people want. Read the story here.

The hidden danger of a successful pitch

Winning a pitch doesn’t tell you much about what it will be like to work with the client who chose you. An amusing Substack essay by Andrew Boulton recounts an agency “win” that quickly exposed how little selection processes reveal about power dynamics, expectations or day-to-day reality. It’s funny in a bleak, recognisable way, but the point underneath is serious: chemistry in a pitch doesn’t necessarily lead to amicable client relations. Worth reading if you’ve ever felt uneasy the moment the congratulations rolled in. Check it out here.

AI prompt of the week: building your investor-specific pitch

Most founders create one generic pitch deck and fire it at every investor, wondering why traction is slow. The truth is that successful fundraising isn't about having a great business – it's about demonstrating why your specific business fits a specific investor's thesis, portfolio strategy and track record. Tailoring your investment case to each target isn't extra work; it's the difference between "interesting" and "let's talk terms".

Why would [Investor's Name or Investor Type] invest in [Company Name] in the [specific industry or market]? Outline the compelling reasons, focusing on the unique strengths, market potential, innovative products or services and the strategic advantages [Company Name] offers. Provide a detailed analysis that highlights how these factors align with [Investor's Name or Investor Type]'s investment criteria and track record.

The anatomy of a deal

Headline price rarely tells the full story in a sale. How and when you get paid can matter as much as the number itself. This visual breaks down the main deal structures founders are likely to encounter and the trade-offs hidden inside each one.

Drop me a line

That’s all for this week – thanks for spending some time with me. As always, my email is open for interesting conversations. See you next Sunday.

Cheers!

Adam

Join the Exit Mode Masterclass!

​Sign up for free and learn how to get exit ready in just six months

​
​Unsubscribe · LinkedIn​