When AI underperforms, the problem is usually upstream.
Redpoint finds what’s really holding your AI back and gets it delivering the business result you funded, usually without starting over. 30+ years of voice, conversational AI and enterprise products at Amazon, AWS, Microsoft, O’Reilly and Collibra.
- Led the testing behind California’s AskCA, built for 39 million residents; Phillip quoted in the Governor’s launch
- A failing automation project turned around in 60 days, exceeding its goal
- Product experience rebuilt as ARR doubled past $200M
- 2000% growth in active Alexa skills
- US Digital Response · State of CaliforniaSafe, effective AI-powered assistance for 39 million US residents
- Speechicon · Voice commerceArchitecting a much loved retailer’s high-volume voice commerce agent
- Cole Insurance · through RedpointBuilt a quoting engine to speed up a specialty insurance agency
What leaders bring us
AI is costing more than what it replaced
Calls and cases are leaking out before the AI can resolve them.
Field note: Look upstream first ↗A launch no one wants to stand behind
Nobody tested it against your rules
Field note: Build trust in, test trust out ↗A pilot that impressed the room and stalled in production
The agent was handed the world, not a job
Field note: Give your agent a job, not the world ↗Spend going up, a deadline sliding, and a team that wants to rebuild
It needs examining, not rebuilding
Case study: From 18% to 33%, fixed upstream
How we work together
All engagements →- 1 · Talk
Free AI leadership session
One hour on where your AI investment stands and what would move it.
- 2 · Diagnose
AI performance diagnostic
Find what’s really holding the result back, with a prioritized plan your team can execute.
- 3 · Test
Launch-readiness testing
Know your assistant behaves before customers, or the press, find out it doesn’t. The method behind AskCA.
- 4 · Architect
Agent architecture
An agent with a defined job, the knowledge and guardrails to do it, and a measure of success.
- 5 · Build
Custom AI build
When the answer is software, we build it, delivered in weeks, not quarters.
Selected work
All case studies →Safe, effective AI-powered assistance for 39 million US residents
Led the behavioral testing behind California’s new AI assistant, AskCA, as lead volunteer consultant: 129 simulated-resident conversations, ordinary and adversarial, with every answer scored against AskCA’s own instructions and the 5 high-stakes checks scored by judges from 3 AI model families. Quoted in the Governor’s launch.
How we tested it: Build trust in, test trust out ↗Architecting a much loved retailer’s high-volume voice commerce agent
A voice commerce agent for loyal, particular callers: flexible, natural conversations, scoped, prototyped and tested before callers meet it.
Built a quoting engine to speed up a specialty insurance agency
Coverage checked, priced from the agency’s own rates, application filled and queued for a person to approve. AI working for an everyday business.
Field notes on making AI work for real.
Build trust in, test trust out
How do you know an AI assistant is ready for 39 million people?
Look upstream first
Two weeks after launch, a home warranty company was ready to pull the plug on its new phone assistant.
Words are still not enough
A few years ago I wrote about three words: “put that there.”
Give your agent a job, not the world
Your AI agent is going off script. Wrong answers. Hallucinations. Inconsistent behavior.
Phillip Hunter
Founder and principalFor 30+ years I’ve designed the places where people meet new technology: voice systems in the 1990s, SaaS in the 2000s, Alexa and Cortana in the 2010s, AI agents now. Redpoint is how I take that to teams whose AI isn’t delivering. Most recently: leading the behavioral testing of California’s AskCA assistant as a volunteer with US Digital Response, architecting a much loved retailer’s high-volume voice commerce agent through Speechicon, and a quoting and pricing engine for Cole Insurance. The case studies on this site are my work.
- Senior Director of Product Design, O’Reilly · Head of Product and Services UX, Amazon Alexa · UX Manager, AWS
- Earlier: VP of Design, SpeechCycle · Senior Principal Designer, [24]7.ai · design lead, Intervoice
- Agentic and conversational AI, speech, complex systems, B2B and SaaS product design
- Speaker at AWS re:Invent, UXLx, IxDA, SpeechTek · Leading by Design Fellowship, California College of the Arts
5 Human Tools for AI Success
Five things you can do this week to improve your AI project.