How Cohort 1 moved from AI-curious to AI-capable — measured against their own April baseline, in their own words. Prepared for ElectraNet Leadership by AI² Solutions.
In April 2026, twenty ElectraNet team members told us how ready they felt for AI. The average sat between cautious and curious. Ten weeks later, Cohort 1 — drawn from Legal, Assurance, HSE, Business Support and Business Continuity — completed an intensive, human-led Copilot programme. Every measured dimension of capability rose, and the distribution of confidence shifted decisively to the top end.
This was not a technology rollout dressed up as training. It was a people-focused transformation: small enough to coach individually, structured enough to measure, and deliberately framed around judgement — when AI is the right tool, and when it isn't. The wins the cohort reports are operational (policy reviews in minutes, meetings drafted from recordings, contract first-drafts off a structured starting point) but the deeper win is cultural: the most common baseline fear — that AI would replace people — was directly reversed in the cohort's own words.
The recommendation to the Board is straightforward. Scale this model. Cohort 1 has flagged the practical blockers — IT governance throughput, branded output templates, human-in-the-loop assurance — and these are addressable. The opportunity is roughly ~3,290 hours per year recovered across a single cohort of eleven, with the cultural conditions for safe, judgement-led adoption already established.
From process clarity, to trusted information, to AI-accelerated execution.
Understand how work happens today — map the process, people impacts, bottlenecks, risks, resistance points and change readiness. Value Stream Mapping exposes waste, duplication, delays, manual effort and the real opportunities to improve.
Define the business value, ROI, governance model, ownership and priority improvement areas. Establish a single point of truth across SharePoint, OpenText, Teams, Dataverse and core systems so information is trusted, findable, secure and usable.
Build the roadmap, assign ownership, simplify workflows, train users and measure outcomes. AI is the accelerator, not the engine — the engine is clear process, trusted data, strong governance and people adoption.
ElectraNet's Cohort 1 was coached individually, not enrolled in a course. The programme started with each person's baseline, was paced around real ElectraNet work, and was framed at every step by the same question: “Is AI the right tool for this task?” What changed for the cohort wasn't the tools — it was their judgement.
Every measured dimension rose. Hover any bar for detail, and switch between raw scores and the percentage change.
It isn't just the average creeping up — the whole distribution shifted. At baseline 45% sat at neutral or below on their understanding of AI. By the end of Cohort 1 that dropped to 18%, and more than half are now very confident.
| Understanding of AI | Baseline | Cohort 1 |
|---|---|---|
| At least “somewhat confident” | 55% | 82% |
| “Very confident” | 10% | 55% |
| Neutral or below | 45% | 18% |
Mean understanding score rose +24.6% (3.50 → 4.36 on a 1–5 scale).
Participants were asked how many hours per day they believe Copilot saves them. The average across Cohort 1 was 1.30 hours per day (median 1.0; range 0.75–2.5). Hover each bar to see how many people reported it.
| Time saved | Per person | Across Cohort 1 (n=11) |
|---|---|---|
| Per day | ~1.30 hrs | ~14.3 hrs |
| Per working week | ~6.5 hrs | ~71.5 hrs |
| Per year (estimate)* | ~299 hrs ≈ 40 days | ~3,290 hrs ≈ 439 days |
| FTE equivalent** | ~0.15 FTE | ~1.66 FTE |
| Indicative $ value*** | ~$14,950 / yr | ~$164,500 / yr |
*Annual figures extrapolated from self-reported daily savings (5-day week, ~46 working weeks, 7.5-hour day). **FTE = 38 hrs/week × 52 weeks = 1,976 hrs/yr. ***Indicative dollar value at a blended $50/hr loaded rate; presented as redeployed capacity, not a cash saving.
Model the impact at any scale. Adjust the cohort size, time saved and blended rate to see the capacity unlocked and indicative annual value. Defaults reflect Cohort 1 assumptions.
Inputs are self-serve assumptions, not commitments. Value is indicative redeployed capacity at the blended hourly rate — not a cash saving. FTE uses 38 hrs/week × 52 weeks = 1,976 hrs/yr.
Pulled verbatim from Cohort 1 and Pre-Champion survey responses. These are the wins the cohort chose to surface themselves — fear reversed, judgement sharpened, time recovered.
My biggest fear was that AI would take over my role and I would lose my job. Since doing the AI training, I have come to realise that it's simply a tool to assist me in doing my role more effectively.
Reviewing a policy against legislation and updating to be consistent with it — all in 5 mins.
Copilot training reinforced disciplined use of AI by embedding a critical upfront question: 'Is AI the right tool for this task?' It sharpened my scepticism into a strength.
I've been able to implement AI into my daily tasks — arranging meetings with multiple attendees simply by asking AI to review Calendar, find the most suitable time and send the invite. This has saved me valuable time.
Tasks such as reviewing clauses, preparing emails, and generating briefing notes now start from a structured first draft rather than a blank page. I focus on refining outputs and applying legal judgement rather than building content from scratch.
I'd like to automate reviewing policies and procedures, and automate tasks that don't require 'thought' — for example ordering corporate workwear or occupied site inspections.
Willingness to actively support broader AI adoption rose from 55% to 70%. And the most common baseline fear — that AI would replace people — was directly reversed.
“My biggest fear was that AI would take over my role and I would lose my job. Since doing the AI training, I have come to realise that it's simply a tool to assist me in doing my role more effectively.”
Cohort 1 translated the training directly into day-to-day practice across HSE, Legal, Assurance and Business Support.
Reviewing a policy against legislation and updating it for consistency — “all in 5 mins.”
Advanced drafting, clause redlining and comparative analysis across complex legal documents, plus stakeholder-ready emails, executive summaries and briefing notes.
AI reviews calendars, finds the best time and sends invites, then drafts the minutes from the recorded session.
Collating data into a functional dashboard that can be adapted into a reusable agent tool for future use.
Faster document search, first-draft generation and proofreading across routine work.
“Copilot training reaffirms that whilst AI can be capable of many things, not all problems require a tech solution.”
The cohort was candid about what would help them go further. These are the practical blockers to scaling beyond Cohort 1 — every one of them is addressable.
Approvals to build agents are “taking a fair bit of time”, with silos between teams slowing momentum.
Lack of branded templates in Copilot forces manual reformatting, eroding some of the time saved.
Participants themselves want guardrails so AI augments rather than replaces judgement.
Appetite for human-in-the-loop review, audit trails and clear data-access controls — especially in Legal and Assurance.
Baseline: the AI Awareness & Usage Baseline Survey (n=20), captured 7–8 April 2026 as the engagement began.
End of cohort — combined Cohort 1: the post-training survey responses taken together (n=11), forming the complete Cohort 1 end-state dataset across Legal, Assurance, HSE and LRG.
