ExplainerAIStrategy

AI Quick Wins: Five Rules for a Year That Proves Value

Start with documents and data, move to assistants and dashboards, and judge the year on outcomes rather than usage.

OMBIRO4 min read
The claimDoes the evidence support it?Each dot: one real test

Satya Nadella, Microsoft’s chief executive, told the World Economic Forum in January 2026 that the AI industry will “quickly lose even the social permission” to use scarce energy unless it improves real outcomes in health, education and competitiveness.1 The same test applies inside an organisation. Staff, boards and funders extend permission to AI only for as long as it visibly improves results. OpenAI’s finance chief, Sarah Friar, made the practical version of the point, calling 2026 the year of “practical adoption”.2

Quick wins are where that permission is earned or lost. Tasks that are too ambitious stall the pilot. Tasks that are trivial change nothing. BCG’s long-standing rule of thumb is to put about 70 percent of strategic focus on people and processes, 20 percent on technology and 10 percent on algorithms, and its 2025 research found that most roadblocks to value involve people, organisation and processes rather than technology.3 Five rules for the first twelve months follow from that.

1. Start where the tools already are

Documents, spreadsheets and slides come first: summarise, clean, draft. Nothing new has to be bought, and gains show within weeks. The next tier, assistants on approved content such as policy questions, needs light set-up by trained champions, not code. Dashboards and integrated assistants take months and are usually built with a partner and owned by a named executive office.

Exhibit 1Quick wins by tier
WHAT IS REALISTIC, AND WHEN 1 NOW Tools staff already have Summarise long documents Clean, analyse and chart spreadsheet data Turn a paper into slides Draft memos and minutes from transcripts Any staff member 2 IN WEEKS Light set-up, no coding A policy assistant that answers from approved documents A rules-and-thresholds assistant Reminders for open conditions Champions set it up 3 IN MONTHS Built with a partner A portfolio dashboard on core system data Impact data flowing into funder reports Assistants integrated into core workflows An executive office owns it Champions trained in Quarter 1 set up the weeks tier. Tiers are indicative. Author's framework.

Author's framework. Tiers are indicative and depend on groundwork.

2. Pass three tests

A quick win is repetitive, checkable and owned. Repetitive, so the hours saved are real. Checkable, so a person can verify the output in minutes. Owned, so a named person is accountable for the result. Ask each department for its three most repetitive tasks, ranked by hours saved, and pick from the top of the list.

3. Run four quarters

Enable, use, automate, embed. Quarter one is policy, practical sessions, a champion in each department, a ranked task list and a map of gaps in systems, data and skills. Quarter two puts AI to work on documents and data and launches a first assistant. Quarter three automates routine workflows and delivers a first dashboard. Quarter four writes the new ways of working down and agrees the second-year plan.

Exhibit 2The first twelve months
THE FIRST TWELVE MONTHS Q1 ENABLE Usage policy Practical sessions A champion in each department Ranked routine-task list Gap map of systems, data, skills Q2 USE AI on documents and data across teams First assistant live within 60 days Q3 AUTOMATE Reminders and trackers on routine workflows More assistants First dashboard Q4 EMBED New ways of working written down Year 2 plan agreed Scorecard reviewed Judged on the business scorecard, not on tool usage Author's framework.

Author's framework.

4. Put a quality gate on every output

AI makes polished work cheap to produce, and that is the risk. Research by BetterUp Labs and Stanford, published in Harvard Business Review, found that about 40 percent of full-time US employees had received low-substance AI output in the past month, each instance costing about an hour and 56 minutes to decipher or redo.4 The remedy is a rule, not a tool: whoever sends AI output owns it, and checks it first.

5. Judge the year on outcomes

Usage is a diagnostic. Outcomes are the proof: time to decision, share of decisions delivered, value captured and cost-to-income, each measured from a baseline set before launch. In McKinsey’s 2025 survey, the roughly 6 percent of organisations reporting significant value were far more likely to have redesigned workflows, not merely added tools.5

“In a time of global uncertainty, institutions also need a compass.” Sidi Ould Tah, President, African Development Bank Group, May 20266

A scorecard of outcomes is that compass. It tells a board whether the programme is working while there is still time to change course.

Exhibit 3Usage versus outcomes
WHAT TO COUNT, AND WHAT TO PROVE COUNT IT a diagnostic Staff using AI every week Hours released on routine work PROVE IT the outcomes Days from submission to decision Share of decisions delivered Value captured Cost-to-income 40% of US employees received low-substance AI output ("workslop") in one month 1h 56m of rework lost per incident QUALITY GATE Whoever sends AI output owns it, and checks it first. Baselines are set before launch; the executive committee reviews outcomes each quarter. Workslop data: BetterUp Labs and Stanford Social Media Lab, 2025. Author's framework.

Author's framework. Workslop data: BetterUp Labs and Stanford Social Media Lab, 2025.

The first twelve months should end with fewer routine tasks, faster decisions and a scorecard that says so.

What the evidence does not show

The workslop study is a US survey of 1,150 full-time employees and measures perceived rework, not audited cost. McKinsey's figures are self-reported. The BCG rule of thumb is a heuristic, not a measured law.

Nadella and Friar lead organisations that sell and depend on AI, and Ould Tah's remark was made about institutions in general, not AI. They are used as signals of direction and as a framing for measurement, not as evidence that quick wins pay.

Sources

  1. Satya Nadella, remarks at the World Economic Forum, Davos, 20 January 2026, reported by CNBC. cnbc.com
  2. Sarah Friar, as reported by Computing, "Big tech leaders opine on the future of AI", January 2026 (comments by technology leaders around the World Economic Forum in Davos). computing.co.uk
  3. Boston Consulting Group, "The Widening AI Value Gap: Build for the Future 2025", September 2025 (survey of 1,250 firms).
  4. BetterUp Labs and Stanford Social Media Lab, "AI-Generated 'Workslop' Is Destroying Productivity", Harvard Business Review, September 2025 (survey of 1,150 full-time US employees). betterup.com
  5. McKinsey & Company, "The state of AI in 2025: Agents, innovation, and transformation", November 2025. mckinsey.com
  6. African Development Bank Group, "Leaders support African Development Bank President Ould Tah's daring strategic vision for economic transformation", 26 May 2026. allafrica.com

Read more on AI