The AI-native bet that unblocked activation and won a major US customer

• A connection only creates value once it goes live, but signed customers were stalling before they ever got there.
• Each activated connection saves the customer ~100 hours a month of integration upkeep and data integrity, so 2,000 connections stuck short of going live meant real value locked up.
• It hit every customer at the same point in the funnel, and onboarding effort wasn't moving the number.

• The funnel flagged the drop, not the cause. I found it by sitting on the onboarding calls and watching how people actually behaved.
• It was one buried step: an API key deep inside the customer's own HR system, on a screen the help docs never matched. Non-technical users couldn't tell if they were even in the right place.
• Better wording couldn't bridge a screen people didn't understand, so the fix had to point at the exact button on their real screen.

• Two non-technical segments hit the wall hardest: HR admins at customer companies switching on employee-data access, and relationship managers at banks and lenders activating a client's pipeline.
• Neither writes code, and both were one buried API key away from going live. Their job was simple: find the right key inside their own system and go live without IT.

Three ways to fix it, from brute force to AI. Here is how each weighed out.

Two buy-ins had to be earned, not assumed.

• Leadership. A working prototype makes the case better than a deck, so I walked the CXOs through the POC: the impact on the stuck connections, how the flow formed a habit once people used it, and where the same pattern could extend across the portfolio. That put the leadership table behind shipping in a week, and behind treating it as a reusable capability, not a one-funnel patch.
• Engineering. It was the higher-risk path, so I earned it rather than mandated it: I prototyped in a day to de-risk the bet before committing 16 people's time, laid out the trade-offs and the capacity it would free, and brought engineering in to pressure-test it and shape it, so the design was ours, not mine.

• Closed-user-group testing is where the bet paid off. Running the overlay with a small closed group of real users surfaced the problems early, while there was still time to iterate rather than abandon the approach.
• The biggest roadblock was a security caveat. Banks and lenders run managed devices that block Chrome-extension installs, so the extension-based flow risked stalling the whole enterprise rollout.
• An in-app iframe was ruled out. The HR platforms block their pages from being embedded, so I couldn't wrap their screen inside our product.
• The fix was a Picture-in-Picture window off the same backend. It needs no install, cleared the security objection, and made the guided flow work inside locked-down enterprise environments.