Are Agents The Right Question?
The enterprise AI conversation has moved rapidly from Generative AI, to AI agents, and now, to agentic AI.
Much of the conversation has been around “Where can we deploy agents?”. I believe the more important question we should be asking is “Where would greater autonomy create better economic value?”
Four Areas Where Enterprise AI Can Create Value
for what I see, enterprise AI create value at four levels – not necessarily stages of maturity but of different dimensions.
- Task level – AI Assistance where AI helps people perform tasks faster or better – bringing individual productivity.
- Workflow level – AI-Enabled Workflows where AI changes how processes and customer experiences operate – bringing systematic productivity.
- Operations level – Autonomous Decisions and Operations where AI continuously discovers, decides and orchestrates across business operations – bringing organisation-wide adaptive operations and new operational value.
- Economics level – AI-Reimagined Business Models where AI transforms how the business creates and delivers products and services – bringing new economic value.
So, Where Could Business Autonomy Matter Most?
Not every process should become autonomous. Not every decision should be delegated to AI. And not every autonomous process requires agents. Deploying agents does not, by itself, create transformation.
The real question is where greater business autonomy can create economic, operational or strategic value that would be difficult, too slow or too costly to achieve through human-led intervention alone.
Agentic or not, we should start by uncovering the characteristics of the business problem rather than the technology. Let think about:
- Decision horizon: weeks and months, or minutes and hours?
- Environment: relatively stable, or constantly changing?
- Constraints: known, or dynamic?
- Objectives: long-term optimisation, or immediate outcomes?
- Decision frequency: periodic, or potentially continuous?
- Cost of delay: moderate, or significant?
- Interdependencies: limited, or connected across many systems?
These characteristics help identify where greater autonomy could create the better value. These questions lead to three Agentic AI hypotheses I believe are worth exploring.
1st Agentic AI Hypothesis — Autonomous Decisions
AI moves from supporting decisions to continuously sensing, reasoning and acting for business decision-making.
What if AI could continuously make and execute complex decisions?
Let look at the recent news about NATS air-traffic-control system failure in the UK1. While we can imagine that Air traffic failure was avoidable, says transport secretary2, the aftermaths can be much more economically damaging than the number of flights directly cancelled suggests.3,4
ATC system failure → flights stopped/restricted → aircraft and crews displaced → cancellations → passenger rebooking → aircraft/crew positioning → further cancellations/delays → accommodation/meal costs → several days of network disruption.
The economic impact goes beyond cancelled flights to lost revenue, passenger rebooking, crew and aircraft repositioning, reduced aircraft utilisation and disruption to later schedules.
Let rethink: what if — when disruption happens — an agentic decision system could continuously assess the changing situation and autonomously coordinate recovery decisions across the airline coordinating all related subsystems?
Disruption → assess impact → evaluate options → recover → monitor → adapt
An agentic system could continuously optimise the recovery. This agentic decision system follows a decision-making principle across multiple domains that could apply to everyday airline disruption, such as overbooking/capacity optimisation, other operations recovery, revenue and aircraft utilisation management, etc.
Airlines already invest heavily in resilient systems. Looking ahead what most beneficial would be to build a resilient decision system that can continuously sense disruption, evaluate thousands of possible actions, and autonomously execute the optimal recovery strategy.
2nd Agentic AI Hypothesis – Autonomous Operations
AI moves from optimising individual processes to continuously orchestrating an interconnected operating network.
What if AI could continuously reconfigure a manufacturing supply network when supply, demand and external conditions keep changing?
A recent House of Commons Library’s research briefing5 highlights that medicines shortages in the UK have become a chronic and structural challenge. Causes include manufacturing and distribution problems, shortages of raw materials and packaging, changing demand and wider geopolitical factors.
A problem in one part of the network can quickly affect another:
Supply disruption → constrained inputs → manufacturing impact → inventory pressure → allocation decisions → distribution constraints → medicine availability → patient impact.
Let rethink: what if – when supply disruption happens – an agentic AI system could continuously sense these changes across the network, reason about their interdependencies, simulate alternative responses and recommend or execute changes?
It could continuously assess supplier availability, inventory, demand, manufacturing capacity, logistics, cost and criticality, and then to determine how the network should be reconfigured as conditions change.
3rd Agentic AI Hypothesis – Autonomous Business Models
AI moves from optimising how the business operates to continuously reshaping the economics of what the business makes, buys and sells.
What if AI could change the economics of the product lifecycle?
For example, in sustainability, global companies are working towards 2030/2032 and 2050 net-zero commitments, but upstream Scope 3 emissions remain a persistent challenge. A significant share of these emissions is associated with purchased goods, materials, components and upstream activities across the value chain, which sit largely outside an organisation’s direct operational control.
A recent AI company has created a solution for the upcycling of LEGO bricks6, offering an interesting signal of what could be possible. Its efforts to recover and recirculate used bricks showed that how a secondary pool of products can become a new source of economic value.
This thinking could apply across many consumer product industries. Imagine applying it to fashion, where the product itself is the result of a long chain of upstream decisions:
Which supplier? Which material? Newly manufactured or recycled? Which product to make? How much? Where should inventory go? How should it be transported? What should it cost?
Let rethink: what if – when determining all dimensions in the product lifecycle – an agentic system could continuously evaluate carbon, cost, demand, inventory, logistics, regulation, margin and risk to make these decisions and dynamically re-evaluate them as conditions change?
Looking Ahead: Agentic AI Will Change Business Decisions, Operations and Economics
The future of enterprise AI is not about deploying more agents or simply adding agentic AI capabilities.
As the agentic AI hypotheses explored in this article move into scaled operations, enterprises will increasingly face a different challenge. As fleets of increasingly autonomous agents operate across interconnected processes, a new operational discipline will emerge around AgentOps, observability, governance, evaluation and continuous improvement. The technology and capabilities to support this evolution are already emerging.
Now, for enterprise AI:
Start with the value. Discover where autonomy matters. Then design only what is needed to realise it.
LESS IS MORE – BEFORE SCALE.
And as autonomy scales, be prepared for the next question:
“Are we ready to govern, operate and continuously improve fleets of agents within the architecture we have created?“
REFERENCE
- Flights impacted by air traffic control issue ↩︎
- Air traffic failure was avoidable, says transport secretary ↩︎
- More flight cancellations as fallout from air traffic control glitch continues ↩︎
- Air traffic chaos and compensation: ‘Shutdown cost us more than £1,000 ↩︎
- Research Briefing – Medicines shortages ↩︎
- The converted mill that could build up a fortune from old Lego blocks ↩︎