The AI tools are live and the expectations are sky-high. Leadership announces a major AI initiative and suddenly the organization is “AI-enabled”. But you- the engineer, the architect, the tech lead are actually living inside of the implementation.
Support tickets are full of edge cases that an AI model can’t handle, coworkers are routing around the tool, and you’re left fielding questions about why productivity isn’t surging yet. According to Goldman Sachs, there is a 38-point difference between how executives perceive AI’s impact and how frontline workers actually experience it.
What happens when the gap becomes your problem to solve?
Technologists are close enough to the ground truth to see what’s happening and credible enough to be heard when it matters. It’s a burden and an opportunity when:
- Executives are measuring success at the point of deployment.
- True adoption requires workflow redesign, not just access. An available tool is not an integrated tool.
- Feedback doesn’t travel upward and the information loop is broken. Optimism compounds when unchecked.
How can you manage up without burning political capital?
Closing the expectation gap doesn’t mean choosing between honesty and self-preservation.
- Lead with data. Quantify the friction. Track AI output that requires correction, how long it takes, which uses cases are working versus which aren’t. An audit gives you something to work with instead of a complaint.
- Reframe the problem into “Here’s how to make it work.”– No one wants to hear that a multi-million-dollar investment is underperforming. Provide a path forward and frame findings as an implementation gap.
- Get Specific. Identify two or three specific workflows where AI is genuinely helping and a few where it is creating friction. Skepticism is read as resistance.
- Build a feedback channel. Create a lightweight way to capture feedback that isn’t going upward – a slack channel, spreadsheet, etc.
- Connect AI performance to metrics that leadership cares about. Translate findings into meaningful info: story points per sprint, time to resolution, error rates, rework cycles.
- Ask for a defined success metric before the next phase. If rollout is expanding, ask what success looks like in 90 days and how it will be measured. It creates a shared reality.
- Document what’s working. Find where AI is adding value and make it visible.
The Big Picture
MIT Researchers found that 95% of enterprise GenAI projects are yielding zero measurable return. That isn’t a tech verdict, it’s an implementation verdict. Tools exist and the gap is the human layer: workflow design, change management, and setting expectations.
If you can manage this gap, speaking the language of implementation reality and executive aspiration, you’ll define how your organization truly benefits from AI.
Deployment is just the beginning.
