The politics that stall your AI rollout

AI deployment is a political problem

Paul Willmott argues that AI rollouts stall on politics, because redesigning a workflow moves power. A manager who halves her team with AI may see her grade cut at the next review, and no training course will change that incentive. The losses land on a few people who organise to resist, while the gains are spread thinly across many who stay quiet.

In my experience, this is the conversation that makes all of us leaders squirm in our seats, because it means admitting that while implementing AI across an organisation may offer great benefits, such as enabling significant reductions in operational expenses, it also means that some people will be worse off.

Only the CEO can make the difficult decisions on how AI will be considered and implemented across an organisation, as every other executive has their own function that they are trying to protect. We as CEOs need to clarify the policy that will be applied company-wide, to ensure that all of our team members will be on the same page.

A good place to start is a question from an HR point of view: what happens to a manager's grade and pay if she uses AI to do the same work with half the team? When all team members can see what happens when time and expenses are saved across the company, it will help teams understand how to adjust their work towards the direction their company is evolving.

Also worth your time

The AI value curve: The economics of applying AI to existing work, McKinsey. McKinsey's own model puts the value of applying AI to today's work at $2.52 trillion a year, with the best returns from helping expensive experts. A 30% rise in AI costs would cut that by about a third, so ask your CFO to test every AI business case at higher cost, and to plan where freed-up people go before the rollout starts.

What on earth is going on with AI pricing and packaging?, Elena Verna. What does a $20 AI plan buy? Vendors price in credits and tokens that map to no unit of work, which makes it difficult for us to determine the plan's ROI. If you buy AI, ask suppliers to translate their credits into work you recognise before you sign. If you sell AI plans, predictable pricing will be your big advantage.

This Week in Charts: Agents Denied, a16z. Shoppers use AI to research products they've already found, but few let it buy: 42% in a survey a16z cites would not trust AI to make a purchase at all, and big retailers are blocking shopping agents. For our CMOs, the job right now is making sure our products read well when AI is doing the homework.

Your product is not the center of the universe, Andy Peatling. As customers hand more work to their own AI agents, products that expect people to come to them lose ground. Peatling argues for making your product usable from inside whichever agent the customer already uses, and for investing in the data and expertise an agent can't get anywhere else. A great question for your CPO this quarter.

Websites taken down worldwide as critical PHP cyber-security update issued, BBC News. AI tools found two critical flaws in PHP, the language behind much of the web, unnoticed for over a decade. The UK's NCSC advises taking systems offline if they can't be patched, and PHP's own advisory says to update now. PHP runs 70% of the world's websites, which means some of your infrastructure might be vulnerable to this massive exploit.

Is your leadership team ready to light the fire?

Upskill with AI Firestarter bootcamps and workshops, built for leaders who are too busy leading to be experimenting with AI.

As a leader, you set the tone for how AI is used in your organisation.

Until next time,
Ryan Snyder
ryan@aifirestarter.com