Most organisations can tell you how many people have access to AI tools. Far fewer can tell you whether those tools are creating value or quietly creating operational friction.
Executive summary
AI tools are increasingly embedded in the daily operations of small and medium-sized businesses. However, inconsistent and unmanaged AI use can generate hidden operational drag that gradually erodes financial performance.
This briefing presents a scenario-based thought experiment for a 50-person SMB illustrating potential annual labour and direct AI tool costs under conservative, moderate and high-friction assumptions. The model deliberately uses illustrative assumptions, not industry benchmarks or guaranteed outcomes, to help leaders appreciate where operational friction may accumulate. Common sources of drag include rework, manager review burden, duplicated effort and unmanaged AI-related spend.
Training alone rarely eliminates these hidden costs, so leaders should measure and address behavioural patterns through an AI effectiveness baseline.
At the core, AI is not a strategy, it's a tool.
The challenge now is not whether people are using AI, but whether that use is creating leverage or hidden operational friction.
Why AI adoption metrics can be misleading
Many SMB functions, such as customer support, account management and administration, now rely on AI for writing, research and data tasks. Without agreed standards, clear ownership and visibility, AI can multiply operational friction instead of reducing it.
This friction emerges as:
- Repeated rework of AI-generated outputs
- Managers spending growing time reviewing inconsistent or poor-quality AI-assisted work
- Teams duplicating effort, solving similar problems with different AI habits
- Unmanaged spending on licences, API use and personal AI subscriptions
- Fragmented workflows caused by uncontrolled Shadow AI tools
These factors impose labour costs often invisible in headline productivity or financial metrics, but can significantly affect capacity and budgets. Customer experience inconsistencies can further harm renewal and growth.
Leadership often underestimates these hidden costs, assuming AI is inherently a productivity gain rather than a potential operational drag unless actively managed.
AI amplifies existing work patterns but does not replace the need for clear standards, ownership and leadership oversight.
Unmanaged AI use creates operational noise and friction, generating labour drag and unpredictable costs. The commercial impact comes not from the technology itself but from inconsistent behavioural execution.
Scenario model: a 50-person SMB thought experiment
To make hidden costs tangible, the following scenario model explores annual labour drag and direct AI tool spend for a hypothetical 50-person SMB.
Labour drag components per week:
- Rework of AI-generated outputs
- Manager review time on AI-assisted work
- Duplicated effort due to inconsistent workflows
Direct AI tool spend components per month:
- Licences for core AI tools
- API credit consumption
- Personal subscriptions to AI tools outside procurement
- Unmanaged tools and workarounds
Key assumptions, illustrative only:
- 50 employees
- Fully loaded employee cost: £50,000 per year
- Approximate hourly cost: £25
- Working weeks: 48 per year
| Scenario | Weekly labour drag | Monthly direct AI spend | Annual labour drag | Annual direct AI spend | Total annual exposure |
|---|---|---|---|---|---|
| Conservative | 25 hours | £1,200 | 25 x 48 x £25 = £30,000 | £1,200 x 12 = £14,400 | £44,400 |
| Moderate | 60 hours | £3,000 | 60 x 48 x £25 = £72,000 | £3,000 x 12 = £36,000 | £108,000 |
| High friction | 110 hours | £6,000 | 110 x 48 x £25 = £132,000 | £6,000 x 12 = £72,000 | £204,000 |
Conservative scenario
Here, AI use is controlled and mostly aligned with standards. Teams perform limited rework, around 0.5 hours per person weekly. Manager review time is modest at roughly 5 hours per week, with minimal duplicated effort of about 2 hours weekly. AI tool spending is largely procured and measured.
Labour drag approximates half an FTE at £30,000 annually, while AI spend totals £14,400 yearly. The combined financial exposure represents a manageable but visible friction in workflows.
Moderate scenario
Rework grows to 1.2 hours per person weekly. Managers spend 15 hours weekly reviewing AI-assisted outputs. Duplicated effort rises noticeably to 10 hours per week due to fragmented team approaches. Unmanaged tool spend climbs due to personal subscriptions and rising API use.
Labour drag totals around £72,000 annually, or 1.5 FTEs, with direct spend of £36,000, combining to £108,000 exposure. This indicates a meaningful operational burden and financial strain linked to inconsistent AI adoption.
High-friction scenario
Rework averages 2 hours per person weekly. Managers allocate 40 hours weekly to review and correction. Teams duplicate effort extensively at 30 hours weekly, reflecting widespread workflow fragmentation. AI spend includes substantial unsanctioned subscriptions and API overages.
Annual labour drag reaches £132,000, almost 3 FTE equivalents, with direct spend at £72,000, totalling £204,000 in illustrative exposure. This level of friction risks significant capacity loss and financial leakage.
Interpreting the model
This thought experiment is not a universal benchmark or precise forecast. It uses illustrative assumptions to surface the potential scale of hidden costs tied to AI use behaviour patterns common in SMBs.
The purpose of this model is not to prove a specific cost. It’s to help leaders ask better questions.
- How much rework exists today?
- How much management review is occurring?
- How many AI tools are being paid for?
- Where are teams solving the same problem differently?
- What would reducing those frictions by 10% look like?
By linking observable behaviours, rework, review overload, duplication and spend creep to approximate annual financial exposure, leaders gain commercial clarity on a key risk area often overlooked in AI discussions.
The value of reducing AI drag
Even modest improvements in AI use consistency can generate meaningful avoided costs. For example, on total annual exposure:
| Reduction in labour drag | Conservative | Moderate | High friction |
|---|---|---|---|
| 10% | £4,440 | £10,800 | £20,400 |
| 25% | £11,100 | £27,000 | £51,000 |
| 50% | £22,200 | £54,000 | £102,000 |
<p class="table-note"> These figures are illustrative and should be replaced with organisation-specific assumptions before being used for decision-making. </p>
These figures demonstrate the potential financial benefits of targeted management efforts to reduce AI-related labour drag.
Why training alone is not enough
While training improves awareness and baseline skills, it rarely solves the core behavioural and operational issues that drive friction.
Training does not clarify ownership structures or standardise workflows to limit duplicated effort. It rarely reduces manager review burdens without complementary controls. Shadow AI usage and unmanaged tool proliferation often continue unchecked.
Managers require support through clearer standards, better visibility tools and operational controls beyond training to reduce friction sustained by inconsistent AI use.
What an AI effectiveness baseline should measure
An effective baseline encompasses operational and behavioural indicators, including:
- Variability and quality consistency of AI-generated outputs across teams
- Manager time on review and correction of AI-influenced work
- Frequency and scope of rework tied to AI use
- Incidents of duplicated effort due to unaligned AI workflows
- AI tool spend patterns, including unmanaged subscriptions and API credit use
- Presence and impact of Shadow AI on workflow consistency
Regular measurement enables leadership to see hidden risks and prioritise focused improvements.
Practical next steps for SMB leaders
- Conduct a cross-team audit of AI use behaviours and tools to identify inconsistencies and fragmentation
- Engage managers to quantify AI-related review and rework time
- Review AI tool spend, including subscriptions outside standard procurement
- Apply scenario modelling with organisation-specific assumptions to estimate exposure
- Develop an AI effectiveness baseline, tracking key operational metrics around AI usage
- Supplement training with clear standards, manager enablement and workflow clarity measures
- Favour continuous operational insight over one-off AI rollouts or compliance checks
Closing statement
The challenge is no longer whether people are using AI. The challenge is whether AI use is creating leverage or hidden operational drag.
Responsible AI adoption demands measured management of behaviours, workflows and costs. Leaders who understand and act on these hidden frictions stand to protect capacity, cash flow and customer experience in an AI-enabled world.
Related
WAIA supports organisations seeking greater visibility into workplace AI adoption, behaviour and operational risk.