Artifical Intelligence and Transformation

Brief a consultant or start with a natural language chat prompt?

Hi there,

I didn’t plan for this Substack to turn almost entirely into an AI realism channel but that’s where I find my thinking going. There was so much noise and market volatility that a number of cool announcements in the AI space were missed: Firebase Studio from Google, which lets you build an application from a natural language prompt and deploy it, for example.

I also got access to Manus - and it is as good as the hype. My focus is vibe coding a risk tool in Windsurf but hopefully next week I’ll get to do more testing of Manus.

This week I’m going to write about the old approach to business transformation, how AI will disrupt this, and what this means for the future of business.

Part 1: The Old Approach to Transformation

How does a new CEO start a transformation? After a period of listening and learning, they typically arrive at a new vision for the organisation. They may have even been given a strategy by the board to implement and have no control over this vision.

In the corporate world, a target operating model is designed from the top down. It gets workshopped with stakeholders, signed off by the board of directors, and through this process, it’s more likely than not that teams of consultants and contractors will be helping turn this vision into reality.

This new target operating model sets out the capabilities, functions, processes, platforms and people required to make the vision turn into a reality. It can start at the 30,000 ft level but at some point needs to drop down into the ugly nature of technical debt, process bottlenecks and people problems.

A top-down approach driven by powerpoint and personal relationship building can often lead to good results, but more often than not, big restructuring efforts and transformations take time and cost a lot of money.

Transforming an organisation in the traditional manner is a lot of hard work and takes time to get it right. In the face of client demands, regulator demands, stakeholder demands and things like macroeconomic headwinds and legislative changes, an awful lot of time is consumed between ideation and execution.

Part 2: AI as a Disruptor of Transformation

My theory is that as the capabilities of AI agents evolve, the traditional transformation and organisation redesign process will change drastically.

Imagine if - instead of engaging consultants or running big workshops for weeks on end - a CEO types a natural language prompt into an AI agent.

What’s the safest way to cut my FTE cost by the end of the quarter?

This agent goes off and chats to the finance agent, the operations agent, the client service agent, the risk agent and the product agent and arrives at a list of recommendations within seconds.

Your safest options are:

1. Let 12 contractor roles in technology expire. Their delivery pace is 41% below expected and this project portfolio is uneconomic with your change in strategy.

2. Combine Team A and Team D in operations. They service the same clients and have 41 duplicated processes that can further be automated.

3. Reduce 21 out of 25 currently open job vacancies. There is already sufficient coverage in these functions inside the risk appetite of the firm.

The CEO then types:

Implement all three options. Let me know when you’re done.

The ability to write in plain English and get systematic outcomes is getting easier and easier with agentic AI. For now, much of the focus is on using these tools in research and software development. But embedding them into operations and finance workflows will become more commonplace.

With the rise of AI, an agentic target operating model will need to be designed. Automation of analysis, modelling, research and strategy will become commonplace as tasks are abstracted from complexity back up to natural language prompts and interrogation through asking questions and examining assumptions.

A more agile, flexible and responsive organisation structure could emerge where a few senior people are using the leverage granted to them by AI tooling to deliver material outcomes and deliver value faster.

For example, the rapid progression of AI image generation capability brings into question the need for enormous teams of designers and creatives when business people paying for the outcomes want to cut costs however they can.

Do you really think the same people who outsourced and offshored everything they could will blink twice before slashing jobs and entire functions if they can be replaced with “good enough” agentic workflows?

The final boss level of agentic AI will be all the agents just talking amongst themselves and optimising how to meet client demands, locate manufacturing plants, interact with regulators, and manage capital and liquidity constraints - like the AI-2027 predictions I shared last week are talking about.

Part 3: Implications for the future of business

What does this type of outcome mean for organisations when they want to transform? I think it means they’ll need to move even faster than they already are. LLM models are already getting so much better every few months it’s astonishing that there are still people bearish on the prospects of the technology.

Any process or procedure that slows down innovation increases the likelihood that a challenger who doesn’t care about any of that eats your lunch.

At the moment, a lot of this innovation is heavily technology focused - just look at the rise of agentic coding tools like Cursor and Windsurf.

But as the level of capability of these tools grows, more lower-level roles will become harder to justify. When Deep Research exists, why do you need lots of juniors writing and analysing stuff? If the first versions are this good, how good will these tools be in 6 months? A year from now?

When agentic tools exist, why do you need enormous amounts of legacy software performing step-by-step processing tasks? You’ll need a risk appetite and AI safety guardrails - but your technology teams will be years ahead of regulators on this given how behind they already are.

It’s increasingly clear that being able to implement an end-to-end project using AI-first thinking will become a valuable skill. I think that regulators will eventually come around to this way of thinking too - instead of wanting reports and checklists, they will start saying “as part of having a banking license, our banking supervision agent will now be fully connected to your technology stack with read only access”.

There will be a lot of problems and tradeoffs and scandals that arise over the next few years - but leaders who can work with AI tooling to get from idea to working implementation at pace will create new operating models that slash costs.

Even before these trends exploded, massive job cuts at Twitter/X that led to the platform still functioning and shipping more functionality than it ever did previously, was heralded by many in the corporate world I speak to as the writing on the wall for the professional managerial class working knowledge jobs.

The rise of agentic models is just the nail in the coffin. At some point in the next few years, a forward-thinking firm in a boring industry will turn out to be earning supernormal margins because they let agentic AI loose on optimising their entire operating model in the name of efficiency - we haven’t seen anything yet.

How can you prepare? Try out everything AI that you can. Think about how every rule in your organisation that you think keeps you safe is hurting your team’s ability to innovate. Think about what needs to be cut - not what needs to be added.

Think about how customers will start using AI tools to take back control of what they previously outsourced or offshored as agentic AI makes “rolling your own” solutions so much cheaper than paying a fortune in technology vendor and change costs.

Over at Global Custody Pro this week, I wrote about the impact of market volatility like we saw in the wake of the Trump tariff announcements on clearinghouses and margin.

Please leave a comment or share this with someone you think might find it interesting. I’d love to hear your feedback.

Regards,

Brennan