Three Questions for Deciding Where to Prioritize AI Agents

Prioritize AI agents by focusing on why customers choose you, how your business makes money, and what constrains the delivery of value.

Connections

AI requires decision-makers to make more decisions without much context. Do we deploy agentic AI? Where? For what? Can AI help improve what matters to the business? How do we show return on investment? All of this happens under pressure from the board, peers, vendors and media. FOMO (fear of missing out) kicks in. You start doubting your views: “Am I the only one missing something important?”

Yet, today, I need to decide who gets the money, which project moves ahead, and be ready to stand behind these decisions. Wouldn’t it be great to simplify this AI world enough to make better decisions that we can justify? That is what this article offers: some stability in an information tempest, anchoring us in proven business approaches that have survived multiple technological revolutions. As a technologist, I have learned to respect patterns that have survived for generations.

Executive summary

AI should only help improve the critical parts of your business today. Nothing else. If it cannot, do not use it.

Below are three questions to help create an AI decision framework for a company of any size.

1. How do we get customers?

We start by asking what brings customers to us. Michael Treacy and Fred Wiersema’s Value Disciplines framework offers three starting points: operational excellence, customer intimacy and product leadership.

Do customers choose us for price and convenience, for the way we understand and serve them, or for the products we offer? Beyond those three disciplines, we can also ask how much our visibility and availability matter. Are we simply impossible to miss?

LVMH, the luxury group, provides an example of connecting technology to customer experience. Its Omnichannel & Data page describes the ambition as elevating luxury with unforgettable customer experiences, and explicitly includes AI and generative AI in that work.

Whatever your situation, AI should help you strengthen your strengths. Nothing else. If it cannot, do not use it.

2. How do we make money?

Every company needs money to operate, including not-for-profits, which need money to make things happen. You make money in different ways.

When I think about making money, I think about high-street shops and plumbers that sell products and services on a one-time basis. This is the same approach supermarkets and consultancy firms use. It has been around forever (and will be around forever).

We also have recurring and access-based revenue. This includes renting assets we own, metered pricing such as cloud services and electricity, subscriptions typical of many SaaS offerings, and leasing.

Marketplaces and platforms can let us increase revenue without a proportional increase in operations. Examples of revenue from facilitating transactions or providing access include transaction fees, IP licensing and advertising.

Finally, we have financial methods: interest income, transaction fees and insurance premiums.

Whatever your situation, AI should help you make more money through your existing business model. Nothing else. If it cannot, do not use it.

3. How do we distribute our offering?

In the end, we need to deliver value to our customers. We build value through a series of steps. Maybe we need to manufacture things, create documents or repair something. Whatever it is, it follows a process, with some steps preceding others and dependencies between them.

Our focus needs to be on the step that limits the performance of the whole process: the bottleneck. Optimising before the bottleneck can mean more items waiting for it to clear. Optimisation after the bottleneck can leave efficient steps running idle while waiting for inputs. There is work involved in discovering the bottleneck. Luckily, in IT, there is a ton of data to help. This is the logic of the theory of constraints.

Whatever your situation, AI should help improve the performance of the bottleneck. Nothing else. If it cannot, do not use it.

Conclusion

When considering AI and agents, tune out the noise. Focus on the here and now. Work with things you understand.

Following this approach gives you a better chance of having an AI success story at the end of the path. It also helps everyone learn what AI can and cannot do, allowing people across the company to come up with new ideas that work.

Questions for leadership discussion

  1. What makes customers choose your organization, and where could AI agents strengthen that advantage?

  2. Which parts of your business have the clearest connection to revenue, and how does that shape where you deploy agents first?

  3. How do you ensure AI agents improve the work that matters, rather than simply automating the work that is easiest?

  4. Where is your organization’s real bottleneck, and what would change for the business if agents helped remove it?

  5. How will you distinguish business impact from improvements that make individual teams look more productive?

  6. What governance model gives agents enough autonomy to deliver value while keeping accountability clear?

  7. How do you balance strengthening today’s business model with exploring opportunities that could replace it?