Leadership, Systems Thinking

MAKE the Call: For Leaders Who Own AI Outcomes

Most leaders are handed AI decisions, not prepared for them. MAKE turns a business owner’s intent into a mandate the team, and the agent, can act on.

Insight

4 min

In Agents Get the Glamour. Foundries Give Them the Firepower. I left a question open. When the builder is an agent, the thresholds belong to the business owner: the HR head, the CFO, the person accountable for the outcome. How do we prepare them for it?

Architecture Is a System of Decisions arrived at the same place from another direction. The ingredient that can change a product’s direction must be owned and driven by the business.

Preparing them starts with a better conversation. I have turned the questions I use into a framework. I call it MAKE.

  • Must achieve: what success looks like

  • Actuality: the realities on the ground

  • Kaleidoscope: the options, explored from every angle

  • Edict: the mandate

I will walk through it with one real case: a supply chain team, a delivery promise, and an agent that was not the first answer.

M: Must achieve

The first question is always the same: What will success look like for this effort?

The answers are usually open-ended. Some leaders are specific and perceptive: they can articulate what they will value if it is delivered, rather than milestones, tactical items, or even technology. Most need direction to get to the root.

The follow-up changes with the answer, but its purpose is always the same: remove the noise.

A supply chain team began with: “We need to use AI to speed up deliveries.”

We asked, So what do you expect to achieve by speeding up delivery?

By the end, the answer was different: “We need to redesign our processes and systems to achieve customer delight. The customer survey says that means a T+2 guarantee of delivery, within budget.”

Notice what moved. AI dropped out of the goal, and the customer came into it.

A: Actuality

Actuality is where the goal meets the ground reality, and it is a rabbit hole if it is not managed well.

The supply chain team had developed tunnel vision, out of the seriousness of the issue and a limited understanding of the ecosystem’s limitations. They wanted T+2 guaranteed, while some of the dependencies would not close in 48 hours.

These blind spots are the greatest risk. A design thinking session helps frame the problem with the realities on the ground. My preferred tool there is the "knowns and unknowns" framework, because it surfaces the built-in assumptions and biases that are often the root cause of a blind spot.

  • Known knowns: what we know and can show.

  • Known unknowns: what we know we still have to find out.

  • Unknown knowns: what we assume without noticing. This is where the biases hide.

  • Unknown unknowns: what nobody has thought of yet.

K: Kaleidoscope

Once the realities are set and the problem is framed, there is a tendency to jump straight to the solution. In this case, the suggestion was agentic AI to reduce routing delay.

Turn the kaleidoscope first. Explore the possible solutions mindfully, from every angle, before choosing one.

In my experience, deterministic systems often prove to be the cheaper and more effective solution, with no need to ground the problem in AI. In other cases, a data pipeline has to be built before anyone should think of a system that orchestrates anything.

For this team, the answer was all of it, in the right order. We redesigned the process to be AI-native. The participating systems were enhanced to provide the right data. The data pipelines were reworked. And then, an agentic AI to orchestrate the process.

The agent was in the answer. It was not the answer.

E: Edict

The last step is the mandate, and I call it an edict on purpose.

The executing team needs clarity of direction, so that architecture, design and the system that becomes part of the ecosystem work in synergy. The mandate cannot be ambiguous, and it cannot leave things to interpretation.

The edict is issued by the business sponsor or owner, not by the coach. It is often shaped consultatively, but it goes out as a clear message, approved and stamped by the sponsor.

“Do what is best for business” is not a mandate. “Get us to T+2 without compromising existing commitments to customers and suppliers, within a budget of $X” is.

We did achieve the T+2 guarantee, within budget.

The edict goes to the team: primarily the people who architect the solution, and through them to the AI that builds the systems and executes the workflow alongside people, depending on the solution. This matters more as the builder changes. An agent, like an engineer, will work to the mandate it is given. An ambiguous edict leaves the door open.

Four questions for your next steering meeting

  • Must achieve: What will success look like, in terms of what you will value, not what you will build?

  • Actuality: What are we assuming that nobody has checked?

  • Kaleidoscope: What is the simplest solution that would work, and do we need AI for it?

  • Edict: Could the executing team act on our mandate without asking what we meant?

The business owner does not need to understand the model. They need to MAKE the call.

CONTINUE EXPLORING

More writing on technology, AI economics, systems thinking, leadership, and mentoring.

CONNECT

Follow the thinking, or start a conversation.

Ideas for building technology systems that create lasting value.

Technology • AI Economics • Systems Thinking • Leadership • Mentoring

© 2026 Prasanna Srinivasa Varadhan. Built as a platform for ideas.

LinkedIn