Across the business world, leaders are under pressure to invest in AI. Buying licences, hiring AI specialists or launching a pilot can be done relatively quickly, but turning that investment into lower costs, faster processes and better decisions is much harder.
I set out in my recent research to find the answer to a simple question: do companies that invest more in AI become more efficient? The answer is more cautious than the current excitement might suggest. On average, AI investment alone had no clear association with increased efficiency. This remained the case when looking for benefits up to five years later.
‘Our rivals are using AI’ is not a business case
That does not mean that AI has no value but that average results hide an important divide. The same AI capability can produce very different outcomes, depending on the organisation. Four factors that affected outcome stand out: competitive pressure, managerial ability, stable ownership and access to long-term finance.
Competition
AI was more strongly linked to efficiency gains in companies facing greater competitive pressure. Where profit margins are tight and customers can switch easily, managers have strong reasons to use AI to eliminate waste, improve scheduling and redesign processes. Businesses with greater market power may have room to tolerate weak implementation or use AI for objectives that do not immediately improve efficiency.
A company cannot choose the intensity of competition in its market, but it can recreate some of that discipline internally. Every AI proposal should identify a specific business problem, a baseline measure and a named person accountable for improvement. ‘Our rivals are using AI’ is not a business case.
Good management
The efficiency gains from AI were stronger in companies led by more capable managers. This is hardly surprising once AI is viewed as an organisational investment rather than a product. A model does not choose the right use case, persuade employees to change workflow or coordinate finance, operations and IT – managers do.
Effective leaders are more likely to direct scarce AI talent towards valuable problems, build the required data and controls, and keep a project moving when early results are uneven. Without that execution, specialists can remain underused and new tools can sit on top of old processes, adding cost without removing work.
Time horizon
Ownership and financing also shaped the research results. The link between AI and efficiency was weaker where institutional investors changed their holdings more rapidly. Pressure for quick results can encourage managers to cut training, shorten pilots or abandon projects before learning has taken place. More stable owners may be better able to tolerate a period of experimentation and disruption.
A multiyear transformation is vulnerable with short-term funding
Companies with more long-term debt also showed a stronger link between AI investment and efficiency. This should not be read as a recommendation to borrow more simply to fund AI. The practical lesson is about matching horizons. A multiyear transformation is vulnerable if it depends on short-term funding or must produce an immediate return. Stable finance gives management time to invest in data, training and process redesign.
Business case
Finance professionals are well placed to bring discipline to AI decisions. They can move the conversation away from how much technology is being bought and towards the value the organisation is equipped to deliver.
A sound appraisal should include the following components:
- Start with outcome, not the tool. Define the process to be improved and measure its current performance. Useful indicators might include cost per accurate transaction, cycle time, forecast error, exception rates or working capital released.
- Count the full cost. Licences and specialist salaries are only the visible elements. Data preparation, system integration, cybersecurity, assurance, staff training, human review and workflow redesign can determine whether the investment succeeds.
- Test the organisation’s ability to execute. Confirm that the project has an accountable executive, a capable operational owner and a cross-functional team with enough authority to change how work is done.
- Fund learning in stages. Use pilots, clear review points and go-or-stop decisions but allow for training and adjustment. Early disruption is not necessarily failure, but repeated failure without learning is.
- Measure value, not adoption. The number of users, prompts, pilots or AI-skilled employees shows activity, not performance. Track operational and financial outcomes against a credible baseline.
- Keep people accountable. AI-supported decisions still require controls, documentation and informed human judgment, especially where errors could affect customers, reporting or regulatory compliance.
The bottom line
The central lesson is not that business should invest less in AI. It is that they must support AI investment with the right people, data, processes and controls. AI cannot compensate for weak management, short-term incentives or underfunded organisational change. In the right setting, however, it can amplify a company’s ability to use its resources well.
For accountants and finance leaders, the most useful question is therefore not ‘How much AI are we buying?’. It is ‘What must change around this investment for it to produce a measurable result?’
More information
Read more findings from the Conditional Gains: When AI Investment Enhances Firm Efficiency research report
Recommended reading
