From Decision to Results

You've identified an opportunity. You've decided to build or buy. Now comes the hard part: actually making it work.

Most AI value is lost in implementation. The technology works; the organization doesn't adapt. This chapter covers how to get from decision to results.

The Implementation Phases

Phase 1: Preparation (2-4 weeks)

What happens before technology work begins.

Define success clearly:

  • What metrics will prove success?
  • What's the baseline today?
  • What target makes this worthwhile?
  • How will we measure?

Write these down. Get stakeholder agreement. You'll need this clarity later.

Establish ownership:

  • Who is the executive sponsor?
  • Who is the project owner?
  • Who are the key stakeholders?
  • What are the decision rights?

One accountable person. Committees don't ship.

Assess readiness:

  • Is data actually available?
  • Are systems accessible for integration?
  • Are users identified and available?
  • Is change management planned?

Surface problems now, not during implementation.

Plan resources:

  • Budget (all phases, including maintenance)
  • People (technical, business, change management)
  • Timeline (realistic, with buffers)
  • Dependencies (systems, approvals, other projects)

Phase 2: Pilot (4-8 weeks)

A controlled test to prove value and learn.

Scope tightly:

  • One process or sub-process
  • One team or location
  • Limited data scope
  • Clear boundaries

The goal is learning, not scale.

Select pilot users carefully:

  • Mix of enthusiasts and skeptics
  • Representative of broader user base
  • Able to provide quality feedback
  • Not so unique that results don't generalize

Build feedback loops:

  • How will users report problems?
  • How will you track usage?
  • How will you measure outcomes?
  • How often will you review?

Document everything:

  • What works
  • What doesn't
  • What's confusing
  • What's missing
  • What the data actually looks like
  • What integration issues arise

Timebox aggressively:

  • Set a firm end date
  • Make a go/no-go decision
  • Don't let pilots drift indefinitely

Pilot exit criteria:

  • Did it achieve target metrics?
  • Is user feedback positive enough?
  • Are technical issues manageable?
  • Is the path to production clear?

Phase 3: Production Preparation (4-8 weeks)

Making it ready for real use.

Harden the system:

  • Handle edge cases identified in pilot
  • Improve error handling
  • Add monitoring and alerting
  • Test at expected scale
  • Security review
  • Compliance review

Build operational capabilities:

  • How is it deployed and updated?
  • How are issues detected and resolved?
  • Who provides support?
  • What's the escalation path?
  • How is performance tracked?

Prepare users:

  • Training materials
  • Training sessions
  • Reference documentation
  • Support channels
  • Champions/super-users identified

Prepare stakeholders:

  • Communication plan
  • Expectation setting
  • Success metrics visibility
  • Feedback channels

Phase 4: Rollout (2-8 weeks depending on scale)

Going live beyond the pilot.

Staged rollout:

  • Start with a subset (another team, region, or use case)
  • Verify in each stage before expanding
  • Have rollback plans
  • Maintain close support

High-touch support:

  • Dedicated support during rollout
  • Fast response to issues
  • Visible leadership attention
  • Frequent check-ins with users

Monitor intensively:

  • Usage metrics
  • Performance metrics
  • Error rates
  • User sentiment
  • Business outcome metrics

Rapid iteration:

  • Fix issues quickly
  • Communicate changes
  • Show responsiveness
  • Build trust through action

Phase 5: Stabilization and Optimization (Ongoing)

From new to normal.

Transition to steady state:

  • Reduce dedicated support
  • Integrate with normal operations
  • Establish regular review cadence
  • Define maintenance ownership

Measure actual impact:

  • Compare to baseline
  • Calculate actual ROI
  • Document for future decisions
  • Share results organizationally

Optimize:

  • Improve based on usage data
  • Address persistent friction
  • Expand capabilities if warranted
  • Retire features that aren't used

Maintain:

  • Monitor for degradation
  • Update models as needed
  • Refresh training
  • Review for relevance

Change Management: The Overlooked Critical Path

Technology implementations fail because organizations don't change. Change management is not optional.

Understanding Resistance

Why people resist AI:

  • Fear of job loss
  • Distrust of AI accuracy
  • Loss of expertise-based identity
  • Comfort with current ways
  • Lack of understanding
  • Bad past experiences with technology
  • Not being consulted

All of these are legitimate. Dismissing concerns doesn't make them go away.

Change Management Principles

Communicate early and often:

  • What's happening and why
  • What it means for individuals
  • What won't change
  • How to provide input
  • What the timeline is

Address job concerns directly:

  • Be honest if roles will change
  • If jobs aren't at risk, say so clearly and repeatedly
  • If they are, be human about it

Involve users in design:

  • Co-creation builds ownership
  • Users know their work better than designers
  • Early input prevents late rejection

Make early wins visible:

  • Quick successes build momentum
  • Share positive outcomes
  • Celebrate adopters

Provide adequate training:

  • Not a single session
  • Hands-on, not just presentation
  • Multiple modalities (video, documentation, live)
  • Ongoing, not just at launch

Support the transition:

  • Help desk available
  • Champions in each team
  • Patience with learning curve
  • No punishment for mistakes

The Adoption Curve

Not everyone adopts at the same pace:

Innovators (first 2-3%): Will try anything. Useful for early feedback, not representative.

Early adopters (next 10-15%): Open to change, influential. Recruit as champions.

Early majority (next 30-35%): Pragmatic. Need proof it works.

Late majority (next 30-35%): Skeptical. Need social proof and pressure.

Laggards (last 15-20%): Resist until unavoidable.

Target early adopters first. Let them influence the majority. Don't fight laggards early.

Common Implementation Mistakes

Moving Too Fast

Symptom: Production deployment before pilot learnings are incorporated.

Consequence: Problems at scale are expensive and visible.

Fix: Patience. Gate progression on readiness, not calendar.

Moving Too Slow

Symptom: Endless piloting, analysis paralysis, perfectionism.

Consequence: Value delayed, momentum lost, opportunity passes.

Fix: Timebox phases. Make decisions with imperfect information. Progress over perfection.

Underinvesting in Change Management

Symptom: Technology works, adoption doesn't.

Consequence: Unused AI delivers no value.

Fix: Budget and staff change management like technical work.

Scope Creep

Symptom: "While we're at it, let's also..."

Consequence: Delayed delivery, diffused focus, failed projects.

Fix: Ruthless scope control. Separate projects for separate goals.

Declaring Victory Too Early

Symptom: Launch celebrated, adoption not tracked, issues not addressed.

Consequence: Slow failure instead of recognized success.

Fix: Measure adoption and outcomes, not just deployment.

Ignoring Feedback

Symptom: Users complain, nothing changes.

Consequence: Users give up, work around the system, reject AI.

Fix: Close feedback loops. Show responsiveness. Iterate visibly.

The Implementation Checklist

Use this for each phase:

Preparation Checklist

  • Success metrics defined and agreed
  • Baseline measured
  • Executive sponsor named
  • Project owner accountable
  • Data availability confirmed
  • Integration path understood
  • Budget allocated (all phases)
  • Timeline realistic (with buffers)
  • Change management planned

Pilot Checklist

  • Scope clearly bounded
  • Users selected (mix of types)
  • Feedback mechanisms ready
  • Measurement in place
  • End date set
  • Exit criteria defined
  • Documentation happening
  • Regular reviews scheduled

Production Prep Checklist

  • Pilot issues addressed
  • System hardened
  • Monitoring in place
  • Support processes ready
  • Training materials complete
  • Training delivered
  • Communication plan ready
  • Rollback plan defined

Rollout Checklist

  • Staged approach planned
  • Support staffed
  • Monitoring active
  • Feedback channels open
  • Leadership visible
  • Quick wins captured
  • Issues addressed rapidly

Stabilization Checklist

  • Adoption measured
  • Outcomes compared to baseline
  • ROI calculated
  • Maintenance ownership clear
  • Regular review cadence set
  • Optimization opportunities identified
  • Results communicated

Building Implementation Capability

Each implementation builds organizational muscle. Capture learning:

After each project:

  • What worked well?
  • What would we do differently?
  • What surprised us?
  • What capabilities did we build?
  • What templates or tools should we reuse?

Build reusable assets:

  • Implementation playbooks
  • Training templates
  • Communication templates
  • Vendor evaluation frameworks
  • Change management approaches

Develop people:

  • Identify emerging AI leaders
  • Create learning paths
  • Build internal community
  • Share knowledge across projects

Implementation is a skill. Like any skill, it improves with practice and reflection.

Building vs BuyingManaging AI Teams and Vendors