Hi Reader

I have been turning over the idea this week that almost nothing costs what it says on the label. Not software, not capital, not the cheap channel that stopped working while nobody was watching. The number you agree at the start is only ever the first instalment, and the second one tends to arrive at the moment you have least room to argue with it.

Enjoy!

The software you have used for fifteen years is not really yours

Richard Haldenby runs Salentis, a UK consultancy of about fifteen people that has used Harvest for time tracking and invoicing for more than fifteen years. This month his monthly bill went from around $130 to $2,110. Harvest was acquired by Bending Spoons in 2025 and has moved from flat per-seat pricing to usage-based pricing keyed to active projects, clients and invoiced revenue. One American customer reported an annual charge climbing from $2,800 to $23,000.

I have some sympathy for the commercial logic: per-seat pricing captures value badly from customers who get a lot out of a product with very few logins. The execution is another matter. A pricing consultant quoted in the piece said Harvest had "completely failed at the transparency test".

The interesting shift is not that a supplier put its prices up. It is that fifteen years of loyalty bought Haldenby no warning and no landing strip. Read it here.

Worth reading if there is a dull, deeply embedded tool in your stack that nobody has looked at since the day it was chosen.

The fastest growing ad channel is also the least memorable

WARC's new commerce media forecast puts global retail media at $200.4bn this year, rising to $223.4bn in 2027, by which point it will account for more than 15 per cent of all advertising. Strip Amazon out and growth falls to 9.8 per cent, the slowest WARC has recorded since it began tracking the sector.

The number that stopped me came from Ipsos research inside the report. Memory encoding for ads on retailer platforms runs 47 per cent lower than for the same ads in ordinary offsite environments, and Amazon, Walmart, Macy's and The Home Depot now average more than twenty ads a page.

So retail media buys the click at the moment of purchase and builds close to nothing you can draw on afterwards. A perfectly reasonable trade, if you know you are making it. Get the report here (paywalled so access through Archive).

For founders selling through Amazon or the grocers, this is the argument for not letting the sponsored listing line become the whole marketing budget by default.

The people your organisation calls difficult may be the ones you need most

Most companies are built to reduce uncertainty. Process, forecasting and hiring for fit all push the same way, and the people who keep questioning assumptions get filed under difficult. That works right up until the environment stops being predictable, at which point the person who irritates everyone by asking why turns out to be the one paying attention.

A friend of mine, Chris Kempt, has written a book about exactly this. The Hunting Party argues that different minds serve different purposes, and that the future belongs to small groups of very different people who know how to hunt together. It is a useful corrective if your last few hires have all thought the way you do.

The Kindle edition is free until the end of Tuesday, read it here.

Worth reading if you have ever described a colleague as hard work when what you actually meant was that they disagreed with you.

A very large advertiser moves its money back out of social

VodafoneThree is putting Three back where it started, aimed squarely at 16 to 29 year olds, with a new platform called "Let Fun In" made with Wonderhood Studios. The hero film features a child ice skating whose costume turns into an inflatable chicken suit, precisely the sort of thing Three used to do before it spent a few years trying to appeal to everyone.

More interesting than the creative was the media plan. Three previously put roughly two thirds of its budget into social. It is now moving back into television, out of home and cinema for reach, with YouTube as the growth channel.

I have written before in this newsletter about the pull of cheap, measurable channels. They look efficient right up until you notice the brand has stopped meaning anything in particular. Read the article here.

For anyone thinking about next year's budget, here is a serious advertiser concluding in public that reach was worth paying for after all.

The moment AI buyers started asking what they were paying for

OpenAI's enterprise revenue has overtaken its consumer business for the first time, roughly two quarters earlier than the company had projected. The annualised run rate is around $40bn, about double a year ago, with business customer numbers up 32 per cent in July alone.

The line that matters is Sarah Friar's description of how buyers have changed. Enterprise customers, the chief financial officer said, have moved "from tokenmaxxing to focusing on cost per unit of intelligence". It is a slightly awful phrase for a genuinely healthy development. Finance directors have stopped funding open-ended experimentation and started asking what a unit of useful output costs.

That discipline arriving from the customer side is far better than it arriving from the vendor side. It is also the point at which AI stops being a slide in the strategy deck and becomes a line in the budget with a number next to it. Find out more here.

Nearly all the talk about AI productivity is about a gain nobody has booked

The St Louis Fed went through 490,000 earnings call transcripts covering 2000 to 2025 and found that 95 per cent of AI-related productivity discussion refers to future gains, against 75 per cent for productivity talk generally. Executives are describing a payoff they expect rather than one that has landed. Utilisation-adjusted total factor productivity, meanwhile, grew 0.07 per cent over the four quarters to March.

There is a second figure I keep returning to. Only 13 per cent of chief HR officers strongly agree their company is AI-ready, against 28 per cent of the wider C-suite. The people closest to how the work actually gets done are less than half as confident as the people describing it from the stage.

I prefer to take an optimistic view. Investment usually does precede measurable return. But it is worth being honest with your board about which side of that gap your numbers currently sit on. Check it out here.

When a lender decides to become the owner

Ares Management is converting its debt position in Charles Taylor, the London insurance services group, into equity and taking control from Lovell Minnick Partners, which bought the business in 2019 at a valuation of around £285m. It follows a rapid break-up: on 12 August Charles Taylor sold its InsureTech arm to Vencora and its US third-party administration business to AvonRisk. What Ares now holds is a slimmer core of claims adjusting and assistance generating roughly £20m of EBITDA.

No price has been disclosed and neither side is commenting. People close to the deal told PitchBook that Ares had already recovered a substantial part of its original investment.

It is a reminder that debt is not neutral capital. Whoever lends you money holds an option on your ownership, and that option only becomes visible when things get difficult. Get the story here.

For founders who funded growth on debt because it looked cheaper than dilution, this is the scenario living in the small print.

Get your own valuation before somebody hands you theirs

Cavendish, the London boutique bank formed from the 2023 merger of Cenkos and finnCap, has hired Teneo to run an independent valuation after unsolicited takeover interest. Chair Lisa Gordon framed it as making sure the board is prepared for any eventuality. The firm turned down an approach from S&W for its M&A advisory arm last year.

I like this rather more than it might first sound. Most boards discover their number in the middle of a live approach, under time pressure, using an adviser appointed in a hurry to somebody else's timetable. Cavendish has done the work while nothing is happening, which means the first credible valuation on the table is its own rather than a bidder's.

For anyone who might sell in the next three years, that sequencing is worth more than any negotiating tactic you will read about. Find out more here.

Eleven applications for every place, and most employers are not opening one

Applications for degree apprenticeships in England have tripled to 21,800 from 8,100 in around three years, while available places have nearly halved to 3,700 from 7,300. Competition for Level 6 places now runs at 11.3 applications per position, up from 2.8 in 2023-24. Searches are up more than 80 per cent this year.

Employers have pulled back because staffing costs and general uncertainty make a multi-year commitment feel risky. That is understandable. It also means the pool has never been less contested.

An SME opening one or two places right now is choosing from a field it could not have competed for two years ago, against employers who have largely stepped away. That is the mirror image of everything else in this issue: cheap because other people are nervous, rather than cheap because it is worth less. Read it here.

AI prompt of the week: supplier dependency review

The Harvest story at the top of this issue is only alarming if you do not already know where your own exposure sits. This one maps the tools, suppliers and platforms your business would struggle to replace, and tells you which to deal with first.

Help me review supplier and platform dependency in my business. My business is [business type], with [number of employees] and [annual revenue]. Our critical tools and suppliers are [list software, platforms, payment providers, logistics partners and any single supplier you rely on]. Our current annual spend with each is [spend per supplier], and we have been with each for [length of relationship].

Create:

A dependency map ranking each supplier by how badly the business would be disrupted if they doubled their price, changed their terms, or removed us tomorrow. Score each one on switching cost, data portability, contractual notice period and whether a credible alternative actually exists.

A shortlist of the three relationships where our exposure is highest relative to what we are getting back, with the specific reason each one is fragile.

For each of those three, a realistic mitigation: what to renegotiate now, what to document or export now, and what a genuine alternative would cost to move to, including the internal time it would take.

Five questions to put to each of those suppliers at the next renewal that would surface a pricing or ownership change before it lands on us.

Base this on how experienced procurement and operations leaders assess concentration risk, not on generic vendor management advice. Be specific about what I should do in the next thirty days, and be blunt about which dependencies are simply the cost of doing business and are not worth fixing.

Drop me a line

Most of this issue is one question in different clothes: what are you paying for something you have not looked at in years? Do reply if you have an answer, or a good story about getting it wrong.

Cheers!

Adam