[2026 kcdc] designing agentic workflows that teams actually use

Speaker: Alexandria Hamilton

See live blog table of contents for more posts


Context

  • One person company
  • 32 agents
  • Bottlenecks with human in the loop. Repeated a number of times that human is the bottleneck. Balance control and speed
  • If fail, because of human
  • “we” – agents part of team

Agents

  • Each agent has own purpose
  • Help with code, calendar, marketing, come up with lyrics, governance, cataloging, checking news
  • Agents doing well so getting promoted to cloud and not human in the loop
  • Org chart of agents; have leads
  • Each agent has a specific thing to do/purpose
  • Skills – ex: firecall, opensco, buffer

When don’t need agents

  • Agents are expensive
  • guardrail: 30 day waiting period before add another agent

Names

  • Some people give them human names and treat as human teammates
  • But want to be able to destroy when don’t need so don’t get too attached

Log

  • Keep log of all workflow
  • Log schedule, handoff, interactions, failures
  • Stored in Obsidian
  • “If it isn’t written down, the AI is improvising it”
  • Helpful for finding old conversations/backup of human memory

Judge

  • The LLM is never the judge
  • Background agents help as judge

Claude

  • Filter ideas with Claude
  • Refine ideas
  • Package and give to agents
  • Using Max subscription. Want to space out token use

Consistency

  • “Consistency is a byproduct of structure, not of intent, memory, or discipline”
  • Slow and steady wins the race
  • Ok if things break. Don’t want agents out in the wild doing whatever
  • Ok with human being the bottleneck
  • Some days failures because human not available
  • Pimon looks for anomalies

Operating Loop

  • Each agent runs certain days/times on a schedule (try to take Sunday off)
  • Want to keep token usage low. Spreading out which days each runs helps
  • Some talk to each other
  • Give each specific things to do

Agent HQ

  • Includes file describing each agent – persona, what does, example invocations
  • Calendar showing when runs
  • Output of Claude code

Health care

  • Marketers need to show accurate
  • Still need to go over data

Relationship with AI

  • Not replacing us, argue with it every day.
  • Partner, not intern
  • Ask Claude how to make improvements: ex to agent schedule

Takeways

  • Agents are good
  • Human is the bottleneck. This is fine. See firsthand everything that went wrong

My take

The session I attended in this room also started 8 minutes late due to video issues. Not the speaker’s fault, but would be nice if someone announced that working on it. Her speaking style is great; high energy and clear which I really needed this morning! I feel inspired. Showing Agent HQ at the end was great!

[2026 kcdc] keep It dull, keep It running: runbooks, routines, and the religion of boring platforms

Speaker: Prerit Munjal

See live blog table of contents for more posts


What want at 2am

  • Consistency
  • Clarity
  • Muscle memory

Groupon’s road to being AI native

  • local experiences marketplace
  • Project Foundry – AI Native operating model
  • One assistant for every employee – Claude. Company default since July 2026
  • 11 people, 12 hours for a migration to a different cloud. 10x lower than years back.

Platform Engineering

  • Don’t need if startup/small company
  • GCP console is good enough
  • Don’t have layers yet
  • DevOps created problems vs solving problem. Now spending more money
  • Platform engineering is everyone’s job but nobody agrees on what it is.

Ai

  • Can’t build the platform for the next 3 years.
  • Know what agents will look like for 6 months

Other problems

  • Don’t want developers creating own account and deploying to it
  • When too many layers, get slower replies
  • Adoption you have to force is a layer you added
  • A platform is a layer unless it removes/replaces one

CI/CD

  • $28K a year for Jenkins
  • GitHub actions much cheaper
  • Developers don’t want to switch because so different
  • Setup agent to write GitHub Actions equivalent, test it and create the PR

Incident

  • Most of an incident is not the fix
  • Times goes to contacting person, gathering and analyzing relevant data
  • Much of time is spent on tasks that can be automated
  • Incidents generally happen a while after deployment because have headroom from when was working
  • Incident is the state a platform is in.

Sample of things to track

  • Deployment activity
  • Who reading PRs
  • Success rate of deployments
  • Who tried to deploy when

Other notes

  • https://herdr.dev – embedded in platform
  • platform vs cli
  • Stateful alerts are moe useful than stateless pings
  • Nobody talks about prompt engineering anymore. Then context engineering. Now harness engineering

Claude

  • Suggested downloading talk, feeding to Claude and pointing it at your codbase

Trusted automation

  • Budgetted
  • Loud by default
  • Exercised on a schedule
  • Degrades to the manual path
  • Distributable

My take

I had trouble paying attention to this session, but not the speakers fault. Today is the 25th anniversary of 9/11 (I’m from NYC). A lot of reflection during the window when everything happened. The content was fine although I would have liked a bigger font (I was in the second row and it was small). That compounded the problem because easier to get distracted.