Craig Horton Advisory
The Transformation Brief
Craig Horton
The Transformation Brief

Good morning. Although politics is not my primary focus, the recent changes to DSIT and the new ministerial appointment prompted me to address the Prime Minister in this week's Brief. Kanishka Narayan has been appointed Minister for Artificial Intelligence, holding positions in both the Cabinet Office and the new business department. He will also attend Cabinet meetings, marking a significant shift in authority.

A quick note from my office in the Netherlands: Burnham's wife, Marie-France van Heel, is Dutch. That made this story more interesting locally than most Westminster reshuffles.

This week, I'm focusing on one question: how is authority shifting inside organisations? The central theme is clear: who has authority, who doesn't, and what happens when AI decisions move faster than accountability. HBR examines workers defending AI decisions they didn't make. Michael Wade and Jochen Wirtz discuss trust and AI-driven growth. We also ask whether AI helps critical thinking or just leads to passive acceptance.

The lead signal

AI Secures a Seat at the Top Table

Andy Burnham became Prime Minister on 20 July. Within 48 hours, the government confirmed that DSIT's functions would be redistributed. The Department for Business and Trade would become the Department for Business, Innovation, Science and Trade. Kanishka Narayan was appointed jointly across that department and the Cabinet Office. He will attend Cabinet.

GOV.UK confirmed both changes this week

Being at Cabinet meetings puts him where priorities, funding, and trade-offs are decided. Most AI roles are still in IT or centres of excellence, usually with limited authority or under a Chief AI Officer without budget control. This gap matters. The issue this week is simple: authority is moving up, but most AI teams still don't have it.

When someone gets a seat at the table, someone else loses influence. AI is a big step forward, even though it's still early days. Organisations need to adapt and move power away from old decision-making structures.

Attending Cabinet meetings gives access. Real authority will show when deciding on AI spending, priorities, and delivery in each department.

From my office

I updated the AI Business Value Framework this week and rebuilt the starting point at aibvf.com/start. Describe one proposed AI initiative as you would to a colleague. The engine resolves the industry and business context, keeps every estimated input visible, then returns Accelerate, Fix, or Stop with its value range, confidence, rules, and audit record.

The work begins after a Fix. The framework now names the plays, owners, stop conditions, and re-score gate. A verdict should change what happens next.

In the retail example, change enablement decides the case. Write the case for change. Name the sponsor coalition. Listen to the people whose work will change. Connect the initiative to one board-level measure. Cut away the scope that does not pay. Then make guardrails and the escalation route visible.

That brings me back to the top table. A score can reveal the gap. Progress still depends on someone with the authority to move money, change controls, and stop spending. The central question remains whether that authority is actually there.

On the radar
1

The people defending the decisions weren't the ones who made them

On 22 July, Harvard Business Review published a multi-year field study by Anne-Sophie Mayer, Elmira van den Broek, and Tomislav Karačić. It examined workers in hiring, lending, healthcare, and public administration. They had to defend algorithmic decisions they didn't make or fully understand. Workers changed recommendations based on their own judgment. They sometimes hid the algorithmic source if it would hurt their credibility. Authority moved to the system, but responsibility stayed with the person facing the client.

Harvard Business Review, 22nd July 2026
2

Responsible AI is now seen as a growth issue

Michael Wade at IMD and Jochen Wirtz at NUS Business School point out that the same tools that allow for personalisation and automation can also lead to surveillance, manipulation, and discrimination. They cite three cases: Apitor was fined for collecting children's location data; Sirius XM Radio faces litigation over a hiring system accused of screening by postcode and alma mater; and Wesfarmers used facial recognition on shoppers without sufficient notice.

Harvard Business Review, July 2026
3

Human reasoning is becoming less reliable

Theodoros Evgeniou at INSEAD, Anne-Laure Fayard at Nova School of Business and Economics, and Melchior Tamisier-Fayard, a future ESSEC doctoral researcher, use research from management, cognitive science, and human-computer interaction.

Harvard Business Review, July 2026

Once again, I appreciate your interest and time to read this brief, and as always, look forward to your feedback and comments. Enjoy your weekend, Craig

The Transformation Brief  ·  brief.craighortonadvisory.com  ·  Archive  ·  Read on LinkedIn