Built Into Work · HR Tech field guide

Find the pain.
Redesign the work.
Put it to work.

A practical guide to a workflow-transformation offsite.

Use a focused day to turn a difficult workflow into a shared design, resolved decisions and an owned implementation plan. The result might be a simpler process, a configuration change, a purchased capability or a custom build. AI is one possible part of the answer.

The output: one bounded workflow, a clear outcome, a tested future-state design, and a named owner for the next implementation step.

What this guide includes

  1. Prepare the work and protect the capacity.
  2. Facilitate a one-day offsite.
  3. Map value and requirements.
  4. Simulate the redesigned workflow.
  5. Commit to a 30-day implementation plan.

Built from Vic Akosile’s HR Tech keynote: The Architecture of AI Adoption. The agenda and worksheets translate the talk’s method into a reusable facilitation guide. Timings and checkpoints are suggested starting points, not a guarantee of delivery.

01 / Before the offsite

Find the pain. Make room to act.

Choose a workflow with a boundary

Start with a trigger and an observable outcome: “accepted offer → employee ready for day one” is more useful than “fix onboarding.” Narrow it further if the owners cannot influence the full journey.

  1. Listen to both sides. Interview people doing the work and people receiving it. Ask for a recent normal case and a case that went wrong.
  2. Bring evidence. Collect anonymized examples of waiting, re-entry, missing information, corrections and exceptions. Separate measured facts from assumptions.
  3. Name the outcome owner. Include a person who can make decisions about the process, plus the people who understand its handoffs.
  4. Bring delivery and control expertise. Invite technical support and relevant data, security, policy or legal owners where the workflow needs them. Use a facilitator and a decision recorder.
  5. Protect implementation time. Agree what existing work will stop, what coverage is needed, or which delivery capacity will be added. Reserve time after the offsite.
Readiness gate: if there is no decision owner, no access to evidence, or no follow-through capacity, use the session for discovery and resolve those gaps before promising implementation.

02 / Suggested one-day agenda

Redesign backward from the outcome.

TimeWork togetherLeave with
09:00–09:30Agree on the outcome, boundary and evidence.Definition of done and baseline.
09:30–10:15Walk through a recent case. Mark waits, handoffs and rework.A current-state map with facts and assumptions distinguished.
10:15–10:30Break.
10:30–11:30Ask what actually has to be true. Separate requirements, deliberate experience choices and inherited habits.Requirements and decisions to challenge.
11:30–12:15Design backward from the desired outcome.A first future-state map.
12:15–13:00Lunch.
13:00–14:00Simulate a normal case, missing information, a changed decision and an exception.Failure points, revised handoffs and owners.
14:00–14:45Choose the first improvement and delivery model.Bounded scope, dependencies and operating owner.
14:45–15:00Break.
15:00–16:00Agree tests, evidence, approvals and implementation capacity.A 30-day plan with names and dates.
16:00–16:30Read back decisions. Resolve ownership. Book the review.An explicit commitment or a clear blocked decision.

Facilitator prompts

“What outcome does this step protect?” “What would break if we removed it?” “Is this a requirement, an intentional choice, or a habit?” “Who decides?” “When someone says ‘then somebody checks it,’ who checks what, and how do they know it is their turn?”

03 / Working template

Define value and what must be true.

Use the seven value dimensions: efficiency, scale, quality, experience, risk, capability and financial impact. Do not treat hours saved as cash savings unless the capacity is actually redeployed or costs change.

04 / Working template

Make the future workflow testable.

For each step, specify the input, action, owner, output and exception. Draw the map on a shared board or repeat the row below.

Trigger / inputAction / decisionOwnerOutput / handoffException / escalation

Where can the four A’s help?

Assistants support a person. Applications provide a purpose-built interface. Automations execute defined steps. Agents can pursue tasks within explicit boundaries. Name the human owner, permitted actions and escalation path for each.

ScenarioWhat to test
Normal caseDoes every handoff reach a named owner? Is completion visible?
Missing informationWho requests it? Can the case safely pause? Who sees the delay?
Changed decisionWhich downstream steps must stop or be updated? Who confirms the change?
Exception or failureWho takes over? What is the fallback? What evidence is retained?

05 / Working template

Put it to work.

Choose delivery capacity appropriate to the problem: self-serve builders, embedded engineers, a central build team, borrowed engineers or a vendor. This is a portfolio of options, not a ladder. The operating owner remains accountable after the builder moves on.

CheckpointCommitment
Within 48 hoursShare the map and decision log. Confirm owners, capacity and the first review.
Week 1Resolve dependencies; validate requirements and baseline; agree acceptance tests.
Weeks 2–3Implement the bounded change. Test permissions, normal cases, exceptions and fallback. Secure the required approvals.
By day 30Review evidence with the process owner. Continue, revise or stop. Confirm ongoing maintenance and the next scope.