From AI Experimentation to Business Value: Why Human-in-the-Loop Still Matters
By Igor Lima
Why trust, governance and human-in-the-loop workflows may be the missing link between AI experimentation and measurable business value.
1. AI is no longer the future
Artificial Intelligence is no longer a future capability reserved for innovation teams and technology enthusiasts.
Research note
Microsoft's Work Trend Index found that 75% of knowledge workers are already using AI at work.
This resonated with me because almost every person and organisation I speak to knows of people using ChatGPT, Copilot or Gemini in some form.
Author's Perspective
The challenge is no longer getting access to AI. Employees are already experimenting with it every day.
2. The gap between experimentation and value
Despite rapid adoption, many organisations continue to struggle to realise measurable business value from AI initiatives.
Research note
The challenge is converting experimentation into trusted, measurable business outcomes.
The issue is rarely access to technology.
3. Trust may be the missing ingredient
The Economist recently explored how leaders should talk about AI with employees.
Research note
Employees are being asked to embrace a technology that causes fear.
It claims leaders are oftening focusing on productivity gains, while employees may worry about job security and relevance.
Author's Perspective
When a bank boss talks about the replacement of “lower-value human capital” by capital investment it highlights one of the biggest challenges in the AI world: trust and adoption.
Successful AI adoption is ultimately a people challenge as much as a technology challenge.
What I see in operational environments
Across customer support, service management and operational teams, the same pattern appears repeatedly.
Information lives everywhere:
- Teams
- Jira
- SharePoint
- CRM systems
- Internal documentation
People spend significant time:
- Searching
- Chasing updates
- Validating information
- Summarising context
The cost is not usually a lack of information. It is the friction involved in finding, validating and acting on it.
The Gartner's warning
Research note
More than 40% of agentic AI projects will be cancelled by the end of 2027 due to unclear business value, escalating costs or inadequate risk controls.
This does not mean AI is failing. It means organisations need clearer ownership, governance, business alignment and risk management.
Why this matters
What caught my attention about Gartner's prediction was not the number itself.
It was the reason behind it.
The technology is not necessarily failing. Many organisations are struggling with governance, ownership, risk management and defining clear business outcomes.
Author's Perspective
Most organisations do not have an AI problem. They have a workflow problem.
The technology is improving rapidly.
What appears to be slowing adoption is often the process around it: ownership, governance, accountability and confidence in the output.
In my experience, people are usually happy to let AI help them gather information. They become far less comfortable when AI starts making decisions without visibility or oversight.
My hypothesis
The biggest barrier to AI success is often not the technology itself, but turning experimentation into trusted, measurable business outcomes — and human-in-the-loop workflows may be one of the most effective ways to achieve that.
Author's Perspective
The most successful AI workflows I have seen do not remove people. They remove friction.
A practical prototype
To explore this idea, I built a small prototype around a common operational scenario: escalation management.
Rather than replacing people, the objective was to reduce the effort required to gather context and prepare decisions.
The prototype asks a simple question:
What if AI gathered information from multiple systems, prepared a recommendation and highlighted risks, while humans retained ownership of the final decision?
AI gathers information from CRM systems, Jira, Teams and knowledge bases, then provides summaries, risk analysis and recommended actions.
Humans remain responsible for reviewing recommendations, approving next steps and handling escalations.
Prototype workflow
Conclusion
The challenge facing most organisations is no longer simply building AI.
Employees are already experimenting with it.
The real challenge is embedding AI into workflows that people trust, organisations can govern and businesses can measure.
Want to explore a similar workflow?
Try the AI demos or explore how practical AI workflows can support operations, knowledge retrieval and human-in-the-loop decision making.