I build agentic voice agents and tune them inside customer production systems. Before that I spent years selling enterprise AI at IBM Watson, Deloitte, Talkdesk, and Observe.AI. On my own time I ship products.
Three products of my own, and three pieces of client work. Open any one for what it does and what I built.
A career-discovery service that walks high-school students from interests to a decision.
A desktop assistant that remembers your work, briefs you on it, and asks before it acts.
Reads project documents, answers questions about them, and tracks the work they describe.
Answers vehicle-protection claims calls for a global insurer and sends each to the right place.
Customer teams arrive with a problem and leave two to four hours later with a working agent.
An agent framework that spots a failed transaction, reasons about it, and proposes the fix.
Small libraries I pulled out of my own production code. Each one fixes a problem I had to solve more than once.
Also public: review-board, audit-ratchet, onet-client, civic-data-adapters, and typesafe-gates.
Most people in this field either build the system or sell it. I do both, and I have done both for over twelve years.
Agentic and voice agents, retrieval systems with guardrails and test harnesses, and the plumbing around them. Python first, then Node.js and TypeScript. LangChain, FastAPI, AWS Lambda, React.
I work inside the customer's own environment and carry an agent from workshop build through pilot to go-live. That includes the dull parts: caller verification, test results, and a runbook for whoever takes over.
Discovery, proofs of concept, RFP answers, statements of work, and the cost case. I closed $30M+ in AI contract value at Talkdesk and about $1M in new recurring revenue in twelve months at Observe.AI.
I built the AI maturity model and the sales-engineering course at Talkdesk, ran worldwide Watson training at IBM, and now run hands-on agent workshops at Genesys.
Four rules, each with the job that taught it to me.
A demo shows what a system can do. A proof of concept on the customer's own calls shows what it will do for them.
At Observe.AI I delivered 10 production proofs of concept. Each one showed a gain on the customer's own measures before the purchase went through.
Decide what the agent must never do before you decide what it should say. Then test for it.
The claims agent can find a contract and route a call. It cannot tell a caller whether they are covered.
A partial fix reported as a partial fix is worth more than a hopeful one reported as done.
Callers were being cut off mid-sentence. I traced it to end-of-speech detection, showed that prompt changes fixed 3 of 4 test calls, and took a timeout fix to engineering.
Slides do not change what a team can do on Monday. Two hours of building does.
Twelve workshops at Genesys, each ending with a working agent the customer's own team built.
My day job is making machines answer. My newsletter, Strange Loops, asks what it would take for one to understand, and what physics says is real.
A series on quantum mechanics and how to read it. More parts are planned.
Copenhagen ruled the question of what the wave function is out of bounds, and the measurement problem stayed unsolved. I see the same habit in how people wave away questions about what goes on inside an AI.
Bohm rebuilt the theory with hidden variables and was mostly ignored. Bell then showed what any realist has to give up.
Everett's theory keeps the mathematics clean and pays for it with a vast number of worlds. I look at what that buys and what it leaves unsolved.
Two direct managers, two senior colleagues, two peers, and a customer, from Talkdesk and IBM Watson. All 42 recommendations are on LinkedIn.
He is super knowledgeable about AI with the special ability to understand customer challenges and present top-notch solutions that combine different AI techniques. Another thing that I like about Michael is his ability to roll up his sleeves and implement complex AI solutions, if necessary. He had a tremendous positive impact on AI sales, showing how to run AI trials successfully (with a very high success rate).
He is the go-to expert whenever specialized knowledge is required, often responding swiftly and effectively, regardless of the time of day. … Mike is exceptionally reliable and dependable, tackling challenges with remarkable ease.
He was part of my Watson core team and was the go to person for AI and Cognitive Search. … He was engaged with customers at world wide level to provide this expertise on complex requirements.
Mike is an asset to any organization. His approach to breaking large, complicated problems into smaller, more manageable problems that the prospect account can resource against is highly valued in today's SAAS market. Combine his problem-solving skills with his background in tech, especially AI, and you'll have an A-Player.
Mike is very personable and I truly enjoyed working closely with him on projects, ideas and initiatives. He was a great person to learn from and be around.
From our very first interactions, it was evident that Michael's priority was always the customer. He has an impressive ability to extract key insights from clients, which he then translates into actionable strategies for implementing AI across their Customer Experience journey.
He ran AI business assessments for some of my accounts and was very knowledgeable about all things automation. He collaborated well in the sales process and showed a keen willingness to understand the customer's business challenges.
Twenty years on one problem: getting a machine to give the right answer. First search, then Watson, now agents.
Forward-Deployed Engineer, AI Mission Control. Earlier: Senior AI Architect, Sales
Build, test, and tune agentic virtual agents in customer production. Before November 2026, led pre-sales architecture for the Genesys AI line.
Senior Sales and Solution Engineer, on contract
Designed an Amazon Q voice agent for a travel customer, on the delivery team.
Architect and Principal Consultant, on contract
Co-designed the ADR framework for back-office exceptions.
Senior Sales Engineer
About $1M in new recurring revenue in twelve months; 10 production proofs of concept across quality scoring, agent assist, and outbound voice agents.
Senior AI Strategic Sales Leader
$30M+ in AI contract value at 135% attainment across 10 straight quarters. Built the AI maturity model and the sales-engineering course.
Principal Architect, then AI Architect on contract
Watson-based voice agents, question answering, and cloud builds, including a voice agent for a California school district.
Specialist Master, AI and ML Solutions
AI strategy and delivery for financial services, utilities, logistics, and healthcare clients.
Worldwide Technical Sales Leader, AI Solutions. Earlier: Technical Support Manager
Worldwide technical sales for Watson, global training, and the last point of escalation.
Senior Director, Global Customer Support and Training
Ran worldwide support for an enterprise search platform and held customer satisfaction above 90%. IBM bought the company in 2012.
MS Physics, University of Pittsburgh
BS Physics, Ohio University
AWS Certified Solutions Architect, Associate
Microsoft Azure: Developer, Solution Architect
Google CCAI
National Merit Scholar
Distinguished Professor Scholarship, Ohio University
Email mpointer@gmail.com, or find me on LinkedIn at linkedin.com/in/michaelpointer.
Text by Claude, voice by ElevenLabs. Both are AI, and answers can be wrong.