Keith Elliott
Founder & CTO, Gittielabs · Author of GittieLabs AgentFlow
Wilmington, Delaware, United States
Actions
Keith Elliott is a technology executive with twenty-five years in software and three CTO roles, who still ships code. He authors GittieLabs AgentFlow, an MIT-licensed multi-agent orchestration framework on PyPI, and has an AI compliance platform running in production at an intelligence-community prime contractor. He publishes applied-AI research at gittielabs.com/research, hosts The Tech Journey podcast, and wrote Swift 3 New Features for Packt. He holds an MBA from Columbia Business School and a BS in Computer Engineering from Georgia Tech.
Area of Expertise
Topics
Why Your AI Pilot Didn't Stick
Enterprise AI programs are usually evaluated on whether the technology works. It almost always does. The interesting question is why capability so often fails to become practice. I'll walk through two modules I built for the same client, on the same platform, delivered within months of each other. One is in daily use: it compresses three to five hours of manual review per case down to about five minutes and lets non-specialists do work that previously required an expert. The other was technically sound, liked by the people who asked for it, and never entered the workflow. The difference had nothing to do with the software. We'll cover the specific conditions that predict adoption — who actually loses time, whose judgment is being replaced versus assisted, where the work sits in someone's day, and what has to be true before a person will trust an output they can't verify. And we'll look at the diagnostic questions worth asking before a line of code gets written, because the cheapest place to discover a workflow won't be adopted is in the first conversation, not after the deployment. Attendees will leave with a practical checklist for evaluating whether a proposed AI workflow will survive contact with the people expected to use it.
Target audience: enterprise technology leaders, CIOs/CTOs, transformation and operations leaders. Preferred duration: 30-45 minutes; also works as a panel contribution. Level: intermediate, non-technical audiences welcome. No special technical requirements.
Context Engineering: Why I Stopped Writing Prompts and Started Designing Systems
Prompt engineering is where most teams start and where a lot of them get stuck. It works until you have several domains, several agents, and a requirement that someone other than the author can understand what the system will do. This talk covers what replaced it in a system I built and now maintain in production. The core move is treating context as an engineering surface rather than a text problem: routing declared as configuration, memory and session scope defined explicitly, tool access permissioned so an agent's reach is a reviewable artifact rather than an emergent property of a long string. I'll cover the architecture concretely — hybrid rule-plus-LLM routing, DAG execution, session memory across multi-step jobs, retrieval scoped per tenant and per user, and page-level citation so a generated claim can be traced to its source. I'll also cover what this made possible operationally: quality gates, human-in-the-loop checkpoints, and per-call cost accounting that took worst-case job cost from over fifty cents to under five. I built this after working in LangChain and LlamaIndex and hitting their limits on real multi-domain workloads, so I'll be specific about where those limits are and when the added machinery is not worth it.
Target audience: AI/ML engineers, platform and infrastructure engineers, engineering leaders building agentic systems. Preferred duration: 30-45 minutes. Level: intermediate to advanced. Reference implementation is open source under MIT (GittieLabs AgentFlow, PyPI). No special technical requirements beyond a projector.
Reading the Evidence on AI and White-Collar Work
There is a confident story circulating about AI and white-collar employment, and a surprising amount of it does not survive contact with the underlying sources. I went through the material most often cited — McKinsey's automation estimates, the Goldman Sachs projections, Anthropic's economic index work, and Bureau of Labor Statistics data — and read what they actually claim rather than how they get quoted. The picture is messier and more useful than the headline version: augmentation dominating replacement in most measured tasks, effects concentrating in specific job components rather than whole occupations, and timelines considerably longer than the discourse implies. This is not a reassurance talk. There are real displacement effects and they land unevenly. But leaders making workforce decisions off the headline version are planning for the wrong thing, and I'll cover what the evidence supports planning for instead. I built an interactive tool for board directors who need to interrogate these claims rather than accept them, and I'll show how to use it.
Target audience: executives, boards, policy audiences, general conference audiences. Preferred duration: 20-30 minutes; also works well as a panel contribution or podcast conversation. Level: non-technical. Sources and the interactive evidence tool are published at gittielabs.com/research.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top