Home / Services / AI Automation Planning

Service Pathway — Emerging

“AI” is being asked to mean four different things.

Most AI initiatives stall because one word is covering everything from a scheduled batch job to an autonomous agent. OnTarget gives organizations a shared vocabulary, a disciplined way to screen ideas, and a realistic roadmap from stalled ambition to working pilots.

Use-case screening · Readiness assessment · Pilot roadmaps · Responsible implementation

The Four Levels of AI Maturity
1

Automation (Rules-Based)

Fixed, repeatable rules — no judgment. Workflow scripting, not AI.

2

Augmented / Predictive

Pattern-based flags and forecasts. A human still acts on it.

3

Generative / Copilot

AI drafts, summarizes, suggests. The human reviews and decides.

4

Agentic

AI decides and acts. Highest payoff, highest risk. Build last.

▲ WITHOUT A SHARED VOCABULARY, PROJECTS GET SCOPED WRONG AND INITIATIVES STALL BEFORE THEY START.

When This Service Is Needed

The signs of a stalled initiative.

AI initiatives rarely die loudly. They stall — in scoping debates, tool evaluations, and pilots that never quite start.

Every team has automation ideas; nobody agrees what 'AI' means for any of them.

A tool decision is being debated before any use case has been selected.

An earlier automation effort stalled and sponsorship has cooled.

The proposed first project is the biggest problem instead of the most provable one.

Ideas are being scoped by enthusiasm rather than criteria.

Governance, risk, and success metrics are afterthoughts to the technology conversation.

Interactive — The Eight Criteria

Is your idea a good automation candidate?

Pick one automation idea your team has proposed and score it honestly. The more “yes” answers, the stronger the Phase 1 fit.

Rule-Based Logic
Can the decision be written as explicit if/then/else steps — no subjective judgment?
High Frequency / Volume
Does it run daily, weekly, or across a large number of transactions?
Standardized Inputs
Are inputs structured and consistent — not free-form?
Low Exception Rate
Are edge cases rare and predictable?
Bounded, Reversible Risk
If it misfires, is the impact limited and easy to correct?
Clear Success Metric
Can you measure the win — hours saved, error rate, turnaround time?
Stable Process
Is the underlying process unlikely to change in 6–12 months?
Accessible Data
Is the data available now, without a major integration lift?
Answer all eight to see where your idea lands.

Delivery Approach

From stalled list to working pilot.

The methodology favors small, provable wins over one large build — sequencing credibility before ambition.

  1. Collect & Screen

    Gather the automation ideas teams have already proposed; score each against the eight criteria.

  2. Rank by Impact vs. Effort

    Prioritize small-footprint, high-frequency candidates — quick wins first, not the biggest problem first.

  3. Tool Decision, Scoped to Winners

    Only after candidates are chosen: build vs. buy for those specific tasks. Off-the-shelf covers most Level 1 needs.

  4. Pilot & Prove

    Implement the top one or two with a before/after metric — hours saved, error rate, turnaround time.

  5. Scale on Evidence

    Use proven results to re-energize sponsorship and open the predictive and copilot wave.

Screening bands: 6–8 yes = Phase 1 automation candidate · 3–5 = predictive/copilot wave · 0–2 = parked.

Specific Deliverables

What engagements produce.

  • Shared AI maturity vocabulary adopted across teams
  • Scored and ranked automation candidate list
  • Build-vs-buy recommendations scoped to selected pilots
  • Pilot implementation roadmap with owners and stage gates
  • Before/after success metrics and measurement plan
  • Governance model for the initiative's next waves

Expected Outcomes

What changes for the program.

  • The initiative restarts with credibility instead of ambition
  • Scoping debates end — the criteria decide
  • First pilots produce provable, sponsor-visible wins
  • Risk stays bounded while the organization learns
  • A realistic path opens toward copilot and agentic maturity
“Favor small, provable wins over one large build.”
The pathway's guiding principle

A single successfully automated process with a clear before/after metric does more to restart an initiative than an ambitious project that takes months to show results.

Relevant Experience

Forged in delivery.

Where the methodology comes from

This methodology was developed in live enterprise operations — customer service and back-office environments where automation ideas were plentiful, sponsorship was fragile, and one more stalled initiative would have ended the conversation. It reflects the same discipline PCM™ brings to large programs: shared vocabulary, evidence over enthusiasm, and momentum built on provable wins.

Common Questions

Frequently asked.

Do we need to buy an AI platform first?

No — that's backwards. Tool decisions come after use-case selection. Most Level 1 candidates need workflow scripting or off-the-shelf RPA, not an AI platform.

Our last automation effort stalled. Is that disqualifying?

It's typical. Stalled initiatives usually failed on vocabulary and sequencing, not feasibility. The screening process is designed specifically to restart them.

What about generative AI and agents?

They're Levels 3 and 4 — real, valuable, and usually not first. The maturity model sequences them after the organization has built delivery evidence and governance at the lower levels.

How does this connect to grid modernization and larger AI programs?

Same discipline at different scale. For utility-scale AI transformation, see the firm's white paper on AI-driven grid modernization — and the stabilization pathway for programs already in motion.

Have a list of ideas and a stalled initiative?

Bring the list. You'll get a screened, ranked view of where to start — and a realistic path to the first provable win. Or start with the white paper: From Automation to AI.