Built, launched and handed over an AI innovation assistant

The situation

Amplify runs an innovation management systems program. Participants work through structured innovation practice, and the questions they ask along the way are predictable in shape and highly specific in content: what does this framework mean here, how does this step apply to their situation, what should they do next. Answering those questions well requires the program's own material rather than a general model's guesses, at a cost that makes sense for a program business.

What we did

The engagement covered the full arc from design to live operation.

• Designed the solution: what the assistant needed to do and, just as important, what it needed to refuse to do, scoped against the program content and real participant questions.
• Built it on Voiceflow, chosen for conversational design control and operating cost: conversation design, knowledge grounding in the program material, and routing to a human where that is the right answer.
• Set up evaluations, so response quality is checked against a standard rather than assessed by impression, and changes ship without guessing whether they made things worse.
• Enabled Amplify's own team to run, update and extend the assistant.
• Maintained and iterated through live use, tuning against what participants actually asked.
• Extended the work into co-creating the AI for Innovation module inside the same program.

What it produced

Ainno launched and runs in production, and Amplify's team owns and operates it. It answers from the program's own material rather than from a general model's guesses, and changes ship against evaluations rather than against impressions, which is the work most assistant projects skip.