Weekly Job Intelligence: The work between AI and adoption

There is a stretch of work between an AI demonstration and an AI system people can rely on. The demonstration shows a capability. The daily workflow determines whether that capability survives conflicting data, exceptions, handoffs, and the judgment of the people doing the work.

That stretch is what Weekly Job Intelligence will study. Hiring signals are one lens into the consequential work organizations are trying to get done: the decisions, roles, processes, and measures that make a result possible. This is not a list of openings. Each week, we will look across reported implementations, research, hiring language, and counterevidence to understand how work is changing. The point is to give leaders and practitioners a more useful question to take back to their own organizations.

This week’s question is whether the process surrounding AI is receiving as much attention as the tool itself.

The friction beneath adoption

A Camunda-commissioned survey (opens in a new tab) of 1,000 process decision makers and 5,000 employees at large organizations in four countries reported that 72% of the decision makers said process challenges had caused AI initiatives to fail. Forty-eight percent said their organizations had rolled back AI use because of negative effects on employees’ ability to work. The employee sample was limited to people who had AI or automation introduced into their day-to-day work. Among them, 44% said they had manually overridden AI output because the underlying process was not set up correctly.

Those are respondents’ accounts in a vendor-sponsored survey, not an independent failure rate for every AI project. Still, the distinction in the responses matters. An employee overriding a result may be exercising good judgment. If overrides become routine because the process cannot handle real cases, “adoption” can hide a great deal of rework.

The Principal Product Strategist, AI Engagement Leader role at Able (opens in a new tab) offers a different kind of evidence. Able describes a leader who will conduct discovery and workflow analysis, assess data and governance, define value and evaluation measures, and stay involved as recommendations become shipped software. A job description is an expression of intent, not proof that a practice has succeeded. It does show the breadth of work one firm believes must sit around AI delivery.

Read together, the survey and posting suggest a hypothesis worth testing in a specific organization: finding another AI use case may be easier than turning an uncertain opportunity into a workable change in how people, systems, and decisions fit together. That includes testing where AI helps, where it introduces another handoff, and where a person must retain authority.

The counterpoint matters. Able also asks for product delivery and technical fluency. Process discovery alone will not ship a reliable product. Nor can one posting establish a general hiring trend. This is a signal to examine, not a claim that the market has settled on a new job title.

A check for the next AI initiative

Before asking whether a team will use an AI tool, map one real piece of work from request to outcome. Where does the information come from? Which exceptions change the path? Who checks a questionable answer? What downstream task waits if an earlier step speeds up? What would count as a better result for both the organization and the person doing the work?

If those answers remain vague, a faster demonstration may simply move the delay. If they are clear, the team can test a smaller intervention, measure its effect, and decide whether the broader workflow is ready to change. That is the work between AI and adoption, and it is where this series will keep looking.

What am I missing? If you have seen an AI initiative succeed or stall because of the surrounding process, leave a comment. Tell me what this reading gets right, where it falls short, or what I should examine next. I’ll review useful questions and corrections as I plan future issues.