When AI Moves From Excitement to Execution

AI is not slowing down. But the conversation around it is changing.

The Australian Financial Review’s look ahead to 2026 captures a clear inflection point. After years of experimentation, pilots, and promise, leaders are now being forced to answer harder questions. Where does AI actually work? Where does it fail? And what does it take to deploy it in environments that involve real customers, real data, and real consequences?

The easy phase is over. What comes next is execution.

The hype cycle is giving way to scrutiny

For the past few years, AI adoption has been driven by possibility. What can models generate? What tasks can be automated? How quickly can costs come out of the system?

Now the focus has shifted. Organizations are discovering that deploying AI in production introduces friction that demos do not show. Outputs need to be validated. Data needs to be protected. Customers notice when experiences feel off, inconsistent, or irrelevant.

The AFR points to early signals of this shift, particularly in consulting, where AI was expected to accelerate delivery and unlock margin. Instead, firms are grappling with governance, oversight, and client concerns around trust and accuracy. The lesson is broader than consulting. AI does not create value simply because it exists. It creates value when it is applied in the right moments, with the right signals, and clear accountability.

This is where many early AI strategies start to break down. Too much focus on capability. Not enough focus on context.

Workforce disruption is real, but incomplete

No part of the AI conversation generates more uncertainty than its impact on work.

Claire Southey, Rokt’s Chief AI Officer, captured this tension clearly in the AFR:

Most major forecasting organisations are predicting very disruptive change to the workforce.

That disruption is already visible. Certain tasks are becoming easier, faster, or automated entirely. But the harder question is what comes next. As Claire noted,  “It’s easy to see the type of jobs that will go away. It’s much harder to predict what the new ones will look like, and the reason for that is all the secondary and tertiary effects of this technology we don’t know yet.”

This uncertainty changes how leaders should think about AI adoption. Treating AI as a straight replacement strategy misses the point. The more durable advantage comes from redesigning how work happens. New roles emerge around validation, orchestration, governance, and optimization, especially when AI is embedded in live customer environments.

Fluency matters. But fluency without a clear place to apply it does not compound.

Put AI where outcomes are visible

One pattern is becoming clear across industries. AI performs best in environments where intent is high, signals are strong, and outcomes are measurable.

In ecommerce, that environment is the Transaction Moment™. It spans from cart to confirmation and represents the point at which customers are actively buying. Decisions made here shape revenue, customer experience, and long-term loyalty in real time.

This is not about adding more content or more choice. It is about making better decisions in the moments that matter most. When AI is applied at the Transaction Moment, feedback loops are immediate. Relevance can be measured. Performance can be improved continuously.

At Rokt, our AI Brain unlocks real-time relevance in the Transaction Moment by determining the next best action and delivering native experiences for each customer. Our global Network provides access to billions of ecommerce Transaction Moments, creating the scale and learning velocity required for AI to improve in production.

This is the practical bar for AI execution. Build where signal quality is high and where value is provable.

Trust is an operating requirement, not a principle

As AI moves closer to customers, trust becomes non-negotiable.

The AFR highlights examples where insufficient oversight led to AI-generated errors, creating downstream risk and rework. In customer-facing environments, the cost of these failures is higher. Inconsistent or irrelevant experiences do not just create operational noise. They erode confidence.

Trust cannot be treated as a promise. It has to be designed into the system. That means:

• Clear thresholds for automation versus human review
• Continuous monitoring of outputs in live environments
• Strong controls around data usage and privacy
• Human accountability for decisions, even when execution is automated
  
  

At Rokt, privacy, control, and transparency are foundational. Our partners maintain full ownership of their data, and relevance is delivered without sharing it. This is critical because customer experience is not something to preserve or protect. It is something to improve. Revenue and profit follow when experiences become more relevant, not when risk is simply avoided.

Expect progress, not instant productivity

There is a temptation to expect immediate productivity gains from AI. History suggests otherwise.

As the AFR notes, when computers entered mainstream business use, it took years before productivity gains showed up in aggregate data. AI is following a similar path. Adoption moves faster than organisational change. The systems need time to mature. Teams need time to learn where AI helps and where it does not.

This does not make AI less important. It makes discipline more important.

The companies that will win in the next phase are not chasing novelty. They are applying AI deliberately, in high-intent moments, with clear measures of success and a willingness to iterate.

What execution-focused leaders do differently

As AI moves from excitement to execution, a few patterns stand out:

1. They start with the moment, not the model.
   Where does relevance change the outcome for the customer?
   
2. They define success in business terms.
   Conversion, value per transaction, retention, and satisfaction, not abstract efficiency.
   
3. They design trust into the system.
   Governance, privacy, and accountability are built in from day one.
   
4. They invest in operating models, not just tools.
   Teams, workflows, and incentives evolve alongside the technology.

AI’s next chapter will not be written by those who talk about it the most. It will be written by those who apply it where it can earn trust, prove value, and improve the customer experience in real time.

In ecommerce, that starts in the Transaction Moment.

‍