In a bold move to turn AI sceptics into daily users, UK law firm Shoosmiths has put a price tag on prompt engineering. The firm announced in April 2025 that if its workforce collectively reaches 1 million Microsoft Copilot prompts within the financial year, a £1 million bonus pool will be unlocked and divided among staff. That works out to roughly four prompts per employee per day — a target the firm says is comfortably achievable with a little habit-building. Shoosmiths is not alone: a growing number of employers are opening their wallets to accelerate adoption of generative AI tools, from cash spot awards to points-based recognition and splashy innovation prizes.

Shoosmiths Wagers £1M on 1 Million AI Prompts

The Shoosmiths scheme is the most eye-catching example yet of a trend sweeping through professional services, pharma, fintech and technology consultancies. The law firm, which rolled out Microsoft 365 Copilot to its 2,500-strong workforce, framed the £1 million pot as a collective challenge. “If each person used Copilot about four times a day the target would be comfortably exceeded,” explained a spokesperson, while stressing that “it’s not just about how many times someone uses AI — it’s about how well we use it and the benefits it will have for our clients.” To reinforce the message, the bonus sits alongside mandatory training sessions, an internal awareness campaign, and a central repository of vetted prompts.

The mechanics are straightforward: the firm tracks aggregated Copilot prompt counts across all employees. If the 1 million threshold is met before the end of the 2025-26 financial year, the pool is triggered. The money is then distributed, likely on a pro-rata salary basis or as a flat payment — details the firm has not fully disclosed — but the overriding goal is to nudge every fee-earner, paralegal and support staff member into forming a daily AI habit.

Beyond the Courtroom: Who Else Is Paying for AI Use?

Shoosmiths may have grabbed the headlines, but reporting by Bloomberg Law reveals a much wider experiment. Sanofi, Brex, IBM and other large employers are piloting their own versions. At some firms, managers hand out spot awards of $100–$300 in cash or vouchers when an employee documents a clear productivity win using AI. Others run point-based recognition schemes where staff can redeem points for gifts, or token programs that highlight clever use cases in company-wide feeds.

Innovation competitions are another lever. Several companies now run quarterly “best use” awards, with prizes of several thousand pounds or dollars going to employees who produce the most valuable, reusable AI application — whether it’s an automated contract redlining tool, a customer-service assistant, or a prompt library that cuts proposal drafting time by 40%. These contests are often judged by a cross-functional panel that evaluates impact, reusability, and compliance with data policies. Leaderboards, both team-based and individual, add a layer of social recognition, signalling that AI fluency is a career asset.

Why Your Boss Thinks You Need a Nudge

The logic behind the payments is disarmingly simple. Organisations have sunk billions into enterprise AI subscriptions and integrations, but the shiny new tools often gather dust. A 2024 Microsoft-commissioned survey found that while 75% of knowledge workers used AI at work, only 39% had received any training — and many still hesitated to trust the output. “Rolling out tools isn’t the same as routine use,” says one HR tech consultant cited in the Bloomberg report. “You have to convert investment into behaviour change, and money is a reliable accelerant.”

Behavioural psychology backs this up. Small, frequent rewards help cement a new habit, lowering the mental activation energy required to type a prompt instead of reaching for an old template. Public awards and leaderboards provide social proof, demonstrating that peers are succeeding and signalling that AI experimentation is safe — provided the right governance is in place. And from a purely financial standpoint, even a six-figure incentive program can pay for itself in months if it unlocks automation that saves hundreds of hours of professional time each month.

The Risk When Usage Trumps Impact

For all the upside, the approach is riddled with pitfalls. The most obvious: rewarding raw prompt counts encourages gaming. Employees might spam the tool with trivial requests simply to inflate the metric. Shoosmiths itself issued a warning about this, reminding staff that outcome quality matters more than volume. Without careful design, a bonus tied to “prompts made” can produce a flood of low-value interactions that do nothing for productivity — and may even degrade the AI’s training signals.

Data security is another red flag. Incentivising heavier AI use increases the likelihood that confidential client data, regulated personal information, or proprietary code will be pasted into a public-model interface. Firms in law, finance and healthcare face regulatory mandates that do not disappear just because a bonus scheme is in place. The UK’s Financial Conduct Authority and the Information Commissioner’s Office have both signalled they will scrutinise how firms govern AI-assisted decisions. If an incentive programme leads to unsupervised, careless prompting, the organisation could be on the hook for a breach.

There are also softer risks: equity. A flat-rate bonus or a simple prompt count favours roles where AI use is obvious — knowledge workers drafting documents or analysing data — while leaving out frontline staff, facilities teams or manufacturing workers. That can breed resentment. Worse, if the tone of rewards shifts from “learn together” to “compete to survive,” it can amplify job anxiety rather than build skills. “Some employees already fear that AI will replace them,” notes a behavioural designer who has consulted on incentive programmes. “Throwing cash at the problem without empathetic framing can make that fear worse.”

Designing Incentives That Work (and What to Avoid)

So what does a sane incentive programme look like? Experts we spoke to returned to three principles: reward impact, not usage; protect data by design; and make sharing mandatory. A good programme might pay a salesperson for publishing a validated, reusable prompt that cuts proposal drafting time by 40% and gets adopted by the wider team. A bad programme simply pays per prompt ingested.

The best designs combine small, frequent spot awards with larger quarterly prizes, and they build in guardrails. Employees must complete a short compliance module before they are eligible; prompts and templates are submitted with a before/after time estimate and a privacy checklist; and winning entries are reviewed by a human panel. One simple template that several consultancies now recommend:

  • Objective: Encourage reusable, high-quality AI assets that reduce task time by at least 20% after peer validation.
  • Eligibility: Any employee who has completed the mandatory one-hour data-handling module.
  • Submission requirements: Short description, time-saved estimate, a two-minute demo video, and a signed privacy checklist.
  • Awards: Monthly spot award of $500 (or equivalent in points); quarterly grand prize of $5,000 plus dedicated time and budget to scale the solution.
  • Guardrails: No confidential data in submissions; rule-breaking entries are ineligible and trigger remediation.

For HR and IT leaders, the playbook is clear: start with a pilot of one or two teams, define baselines, pair incentives with protected learning time (say, four hours a month of “AI sandbox” sessions), and measure success not just by adoption rates but by validated time savings, reuse of assets, and compliance incidents. Typical KPIs include the percentage of the target workforce using approved AI tools weekly, the number of tasks automated or materially improved, and the number of internal templates adopted by multiple teams.

What This Means for Windows Copilot Users

If you are a Windows user in a corporate environment, these trends are likely to reach your desktop soon. Microsoft 365 Copilot — deeply integrated into Word, Excel, Outlook and Teams — is by far the most common enterprise AI tool being incentivised. Expect your employer to announce a training module, a usage dashboard, or even a bonus tied to Copilot adoption. For power users and IT admins, the operational implications are significant: prompt logging and monitoring may become more pervasive, raising questions about employee surveillance and data governance. Admins should ensure that Data Loss Prevention policies are updated to cover Copilot interactions, and that user consent for any monitoring is documented.

As an individual employee, the safest approach is to lean in — but with caution. Complete the training, try out the tool, and document your wins. Avoid pasting any information that you wouldn’t put in a public email. If your firm runs a competition, submitting a simple, sanitised template you genuinely use is a low-risk way to participate. And if you have concerns about how your usage data is tracked, ask your IT or compliance team for a copy of the policy.

The Road Ahead

Cash for Copilot prompts may be the workplace story of 2025, but it is unlikely to be a permanent fixture. Most incentive programmes are designed as short- to medium-term bridges between platform rollout and durable behaviour change. The real test will come when the novelty fades and the prizes dry up. Research on habit formation suggests that external rewards can jump-start a routine, but only if the new behaviour is eventually integrated into job design, performance objectives, and ongoing learning.

Shoosmiths and its peers are betting that a year of nudging will be enough to embed AI into the fabric of daily work. The early data will be instructive. If the experiments produce genuine productivity gains — and if firms manage to avoid the twin traps of gaming and data leakage — expect the practice to spread rapidly. If not, the £1 million pot may be remembered as an expensive lesson in the limits of financial motivation. Windows users and IT pros should watch closely: the results of this grand experiment will shape how — and whether — your own employer pays you to talk to machines.