Mid-market finance thinks it's ahead on AI. The numbers tell a different story
Most mid-market finance leaders believe their teams are leading on AI adoption. Fresh research commissioned by cloud accounting provider iplicit surveyed 250 UK-based finance leaders and found that 83% say they've already adopted AI in their function.
Leaders also rate finance ahead of other departments on this shift: 69% say they're ahead of HR, 67% ahead of sales, 56% ahead of engineering and 55% ahead of IT. Confidence is high-but so are manual workloads.
The pressure cooker: manual work is still eating the week
Over half of finance leaders in smaller SMEs (50-249 headcount) spend 5+ hours a week on manual work. The biggest time sinks are manual data entry (42%), reconciling accounts and transactions (40%), and invoice processing (28%).
If AI is in place, but the workweek still looks like this, adoption isn't solving the right problems-or hasn't made it to core processes.
The adoption gap inside finance
Headline adoption looks strong, but depth is shallow. While 83% claim some level of adoption, only 35% say they've fully adopted AI across finance. Almost half (48%) have only partly adopted it.
The role split is stark. Over half of CFOs (51%) believe their function is fully adopted, compared with just 19% of Financial Controllers. Senior leaders see tools in play. Controllers look at whether the day-to-day has fundamentally changed-and for many, it hasn't.
That same perception gap shows up in department comparisons. A higher share of CFOs and VPs/FDs say finance is significantly ahead of marketing and IT than FCs do. Translation: visibility at the top doesn't equal transformation on the ground.
Use cases: strategy vs operations
AI for strategic decision-making is most popular with CFOs (48%). Integration across functions is also leadership-led: 56% of CFOs and 55% of VPs/FDs cite it, compared with just 16% of FCs.
Leaders see AI helping with board commentary and quick analysis. Controllers see the hours it still takes to prepare clean, reconciled data before any analysis can happen. Today, AI is strong at the presentation layer; core processing is still largely manual.
The confidence and training gap
37% of respondents call themselves "AI champions"-but confidence skews senior: 46% of CFOs and 38% of VPs/FDs versus 26% of FCs. Meanwhile, only 32% have received formal AI training from their employer.
Most are teaching themselves by researching online (60%) or learning by doing (51%). That's a risk. Without training and standards, AI stays as ad hoc experiments, not durable process change.
Where finance is headed next
Leaders plan to expand usage quickly over the next 12 months. Combining current and planned adoption, finance expects to use AI for:
- Expense and spend management: 94%
- Anomaly and fraud detection: 89%
- Forecasting and scenario planning: 85%
The top expected benefits are speed and efficiency (52%), accuracy and error reduction (46%), better workflow coordination (37%) and improved decision-making (35%).
What this means for finance leaders
There's a clear disconnect: leadership sees adoption; operators see partial tools and manual bottlenecks. Deployment and meaningful adoption are not the same thing.
The fix is practical: treat AI as process redesign, not a tool rollout. Start where the hours and errors live.
A 90-day plan to turn AI from hype to throughput
- Map the work: Document month-end, AP, AR, and reconciliation steps. Time them. Identify rework and handoffs. You can't automate what you haven't made explicit.
- Clean the inputs: Standardize vendors, COA, approval paths, and data formats. Automate ingestion (OCR with validation rules) before you automate analysis.
- Pick one high-friction process: For many, that's invoice capture, 3-way match, or bank recs. Define success metrics (cycle time, touch count, exceptions, error rate).
- Pilot, then standardize: Run a 4-6 week pilot with a limited scope. Lock in controls, exception handling, and audit trail. Roll out in waves.
- Train by role: Give CFOs decision-support workflows and FCs operational automations. Provide short, role-based playbooks and office hours.
- Govern from day one: Set data access, model usage rules, and review checkpoints. Use a simple risk register for use cases and vendors.
Metrics that prove real adoption
- Manual hours per week on AP, AR, and reconciliations
- Close cycle time and % of auto-reconciled transactions
- First-pass match rate and exception rate
- Error rate and rework hours post-close
- Policy compliance and audit trail completeness
Governance that keeps you safe
Before scaling, align on risk, controls, and accountability. If you need a simple north star, the NIST AI Risk Management Framework is a practical reference for process owners and IT. It helps you formalize how you evaluate vendors, monitor models, and document decisions.
Training and tooling that actually move the needle
Given the low rate of formal training (32%), closing the skills gap is a fast win. Start with short, practical courses for finance roles and build a common language around data quality, prompts, controls, and exception handling.
Bottom line
Finance is confident-and partially adopted. To convert that into real efficiency, shift focus from dashboards to the plumbing: data capture, reconciliation, approvals, and controls. Start small, standardize fast, measure relentlessly.
If your weekly manual hours aren't dropping, AI isn't "adopted." It's just installed.
Your membership also unlocks:
