IWD 2026
Designing AI Workflows Others can Rely on
Adejoke (JK) Adekunle
Founder, Abby
What's one repetitive task your team does every week that you think should be automated?
A workflow is just how work gets done
What are the design realities?
Built for You
  • You carry the context
  • You decide what good looks like
  • You adjust on the fly
  • You trust it because you built it
Built for Others
  • Requires organisational context
  • "Good enough" is subjective
  • The workflow needs to handle variation
  • Trust needs to be earned
Deciding what to build
Think Like a Product Builder
Not an AI enthusiast.
How to prioritize
Define the business outcome, not the feature.
What it produces ✗
  • "A dashboard"
  • "A summariser"
  • "A report generator"
  • "An AI assistant"
What it solves ✓
  • Stops reps spending 2h on call notes
  • Lets ops catch errors before they ship
  • Means comms never miss critical updates
  • Frees analysts from weekly data pulls
If you can't describe the outcome to a senior leader in one sentence, you're not ready to scale.
Understand your users
Useful to 1–3 people
  • Personal productivity hack
  • High tolerance for rough edges
  • Builder is user, instant feedback
Useful to 10+ people
  • Requires documentation & onboarding
  • Demands consistent quality
  • Needs a named owner(s)
Whose role changes, and what are their needs?
The End User
"Will this make me look bad?"

Needs: confidence signals, escalation path
The Manager
"Who's accountable?"

Needs: ownership clarity, business case
IT / Ops Owner
"is our data protected?"

Needs: documentation, audit trail, support path
Find co-owners
Who owns quality?
A named product owner.
Who handles failures?
Define escalation path before launch.
Who owns maintenance?
Plan for model, prompt, and data drift.
Who approves changes?
Add a lightweight review step for changes.
driving adoption
Getting It to Scale
How to start, involve people, and make it stick.
Find your champion, make them the star.
1
Internal sponsor > internal enthusiast
A senior sponsor's success story opens more doors.
2
Their problem, your solution
Solve a problem they genuinely care about.
3
Ask for their air time
Let them demo and take the credit.
You are the evangelist
Praise others using it
Publicly credit teammates for adoption.
Demo business outcomes
Focus on results, not just AI.
Ask for feedback openly
People support what they helped shape.
Address anxieties
What does this replace?
How steep is the learning curve?
Is there a fallback when it fails?
Keep humans-in-the-loop
Oversight strengthens reliability.
Human review is a designed checkpoint that protects system credibility.
Where to add checkpoints:
High-stakes decisions, customer comms
Low-confidence outputs, add confidence flag
Infrequent edge cases, manual review
🎁 Build your first AI workflow
Build fast
Simple to complex workflows
Automate anything
Data, comms, and more
Scale with ease
Built for any team size
10,000 free credits for your first month.
Coupon: JKSHOP
Not sure where to start with AI workflows?
  • Find the most repetitive task you can hand off to AI
  • Get an instant report
  • Invite your team
  • Compare recommendations

Let's Chat
Adejoke Adekunle
Founder, Abby
adejokeadekunle