Experience across
AI agents and assistants you can watch or try, each with a walkthrough of how it works.
Live demo + sample callsVoice AI agent
Inbound calls pull staff away from higher-value work.
This agent answers routine customer calls, captures each request and checks availability, then flags complex requests for human review with captured context.
Design Principle
Availability comes from your systems, not AI guesswork, and every call ends as a structured record.
Sample call · Escalation
Demo videoMultilingual support assistant
Sale days flood support with the same questions.
This assistant answers in the customer’s own language, with sources, and reuses saved answers to repeated questions.
Design Principle
Search each language separately and reuse saved answers to repeated questions.
Result
Repeat answers 100×+ faster in lab benchmarks.
Controlled AI Lead Triage · A lead-triage workflow connecting form intake, AI interpretation, routing rules, CRM records, and human-review alerts.
See the evidenceAbstracted from real client scenarios and rebuilt as demos to show the design decisions behind each system.
Selected AI systems work across sales, ad review and publishing—with design decisions, evaluation results and supporting materials.
At-a-glance Comparison
Explore the case studies.
| Case | When it helps | How it works | What it demonstrates |
|---|---|---|---|
| Sales Renewal Drafting Assistant | Preparing a customer renewal | Uses earlier deal notes when drafting | Drafting with prior promises and follow-ups in view |
| Ad Launch Decision Gate | Reviewing uncertain ad decisions | Requests evidence and adds checks when needed | Protecting numeric decisions from unsupported changes |
| Newsroom AI Quality Gate | Reviewing AI drafts before publication | Checks required information and references; records reviews | Traceable reviews and repeat checks after fixes |
Keep earlier promises in view when drafting renewals.
Design choice
Keep those details between conversations and check the draft before marking it ready for review.
Result
Drafts passed automated checks in 12 of 12 simulated renewal situations, up from 4 of 12.
Trade-off
Keep context between conversations, at the cost of maintaining and updating deal memory.
Memory lookup and behavioral checks add work before a draft is ready for review.
Check the evidence before changing an ad decision.
Design choice
Reserve extra checks for uncertain cases rather than spending the same effort on every request.
Result
A safeguard reduced mistaken ad blocks from 1 to 0 in an 11-case lab test.
Trade-off
Check uncertain requests more deeply, while relying on lighter checks for routine ones.
Review AI drafts before publication.
Design choice
Keep the AI’s recommendation separate from the decision to release a draft.
Result
Release checks identified all 5 deliberately flawed drafts in a 12-case controlled test.
Trade-off
Hold drafts that fail required checks, even when that adds another review step.
I connect business needs, technical decisions, and team execution—from defining the problem to testing a solution and planning rollout.
Clarify where AI or cloud technology could help—and what is worth testing.
Shape the architecture and work through cost, performance and risk.
Build a focused prototype and test the assumptions that matter.
Bring technical and business teams together around decisions and rollout.
Got something in mind? Let’s talk.
Tell me what you're trying to achieve, your context, and key constraints.
I usually reply within 1–2 business days.