Skip to main content
BlogDigital Transformation & Product Strategy

The boring AI revolution happening in back offices

Digital Transformation Practice3 min read

The AI demos that get the most press are rarely the deployments that pay back. I want to talk about the unglamorous wins. Accounts payable. Contract review. Internal knowledge search. The ones that do not get on the keynote stage but quietly account for most of the productive AI work I see in the market right now. Microsoft and LinkedIn's 2024 Work Trend Index found 75% of knowledge workers using AI at work, with power users saving roughly 30 minutes a day [1]. BCG's 13,000-worker survey put it higher, at five-plus hours a week for regular GenAI users [2]. The interesting thing is where those minutes are coming from.

Take accounts payable. The Hackett Group's 2025 Digital World Class Matrix on AP solutions found leading providers delivering up to 3.5x productivity gains through AI-driven invoice classification, e-invoicing compliance, and approver intelligence [3]. Deloitte's Global Business Services survey found 72% of shared service centres have already implemented robotic process automation, although 56% of them sit at what Deloitte calls medium automation, between 25% and 50% of work automated [4]. That gap between aspiration and execution is most of the story. The tooling is real. Adoption is uneven.

The headcount question is where I get pushed hardest in board meetings. My honest answer is that the changes I observe in the market happen mostly through attrition, not layoffs. Gartner's 2024 finance survey found 58% of finance functions using AI and predicted 90% will deploy at least one AI-enabled solution by 2026, but Gartner also predicted that fewer than 10% of finance functions will see headcount reduction over the same window [5]. The capacity is being reallocated, not retired. That can change at scale, sustained over years. I do not pretend otherwise. The current data does not show it yet.

Contract review is the second pillar. Thomson Reuters Institute's 2024 Future of Professionals report estimated generative AI could save legal professionals four hours a week today and roughly 12 hours a week within five years, with document review the most-cited use case at 77% [6]. The qualifier matters. A 2024 Stanford HAI study found legal-AI tools hallucinating in at least one out of six benchmark queries, even on commercial products marketed for legal research [7]. Translation: the tool tags clauses, surfaces deviations, and drafts first passes. A human lawyer still decides. That is the actual workflow, and the productivity figure assumes it.

Internal knowledge search is the one I think is most undertold. McKinsey's 2023 Economic Potential of Generative AI report estimated knowledge workers spend roughly 1.8 hours a day searching for and gathering information, which is the boring background cost almost no one budgets [8]. The published evidence on what happens when you fix that is encouraging without being miraculous. Brynjolfsson, Li, and Raymond's NBER paper on a generative AI assistant deployed across 5,179 customer support agents found a 14% average productivity gain, climbing to 34% for novices, with near-zero gains for the most experienced staff [9]. A Forrester Total Economic Impact study commissioned by Glean put the upper-bound figure at around 110 hours saved per employee per year via AI-powered enterprise search, which is directional rather than independent, but consistent with the smaller controlled studies [10].

The pattern across all three is the same. The work being automated is high volume, low judgment, and previously consumed time on tasks the humans did not particularly want to be doing. The AI does not handle the hard cases. It handles the easy ones at speed and surfaces the hard ones for human attention. That is not a story about replacement. It is a story about reallocation.

There is a second pattern that is less commented on. The novice-skew finding repeats across studies. McKinsey's 2024 State of AI survey found cost reductions concentrated in supply chain (61%) and service operations (58%), both functions with high junior-to-senior ratios [11]. A controlled 2023 GitHub Copilot experiment by Peng and colleagues had treatment-group developers completing a coding task 55.8% faster than control, with the largest gains for less-experienced programmers [12]. Adoption accelerates when a relatable peer surfaces a visible saving. It rarely accelerates because a CIO sends an email.

If you are a CFO modelling AI productivity for the board, the suggestion is simple. Stop looking at the demo videos. Look at your back office. Pick the highest-volume routine processes in your shared services. Run a 90-day pilot with proper measurement. The win is more likely to look like Bain's 2024 financial-services survey, modest reported productivity boosts in software development, IT, and customer service [13], than like the keynote slide. Unimpressive next to your competitor's announcement. Real, which is an advantage that compounds.

Coderex advises CFOs, COOs, and shared services leaders on the operational shape of these decisions: which back-office processes pilot well in 90 days, what proper baseline measurement looks like, and how to design the rollout so the novice-skew finding actually surfaces wins rather than getting buried by senior-team scepticism.

Expect the 2026 round of CFO surveys from Gartner, Deloitte, and PwC to show finance-function AI adoption past 80% even as headcount reduction stays in the single digits. Expect Hackett's next AP world-class report to push productivity multipliers above the current 3.5x as agentic AI matures past traditional RPA. Expect the legal industry's hallucination problem to drive published vendor-side accuracy improvements before 2027, while the productivity claim continues to assume senior human review.

Methodology note: This article draws on published industry research from McKinsey, BCG, Bain, Deloitte, the Hackett Group, Gartner, Microsoft and LinkedIn, the Thomson Reuters Institute, Stanford HAI, and the NBER Working Paper series, all dated 2023 to 2025. Where vendor-funded studies are cited (Forrester TEI of Glean), this is flagged in the references. I have not surveyed primary client deployments for this piece. Numbers describe the wider market, not specific Coderex engagements.

References

13 sources, all verified at the time of writing

  1. [1]Microsoft and LinkedIn, 2024. 2024 Work Trend Index Annual Report: AI at Work Is Here. Now Comes the Hard Part. Microsoft and LinkedIn. https://news.microsoft.com/source/2024/05/08/microsoft-and-linkedin-release-the-2024-work-trend-index-on-the-state-of-ai-at-work/.
  2. [2]Boston Consulting Group, 2024. AI at Work in 2024: Friend and Foe. Boston Consulting Group. https://www.bcg.com/publications/2024/ai-at-work-friend-foe.
  3. [3]The Hackett Group, 2025. Digital World Class Matrix: 2025 Accounts Payable Perspective Summary Report. The Hackett Group. https://www.thehackettgroup.com/insights/dwc-matrix-accounts-payable-perspective-2511-summary-report/.
  4. [4]Deloitte, 2024. Global Business Services Survey 2024-2025. Deloitte Consulting. https://www.deloitte.com/us/en/services/consulting/services/shared-services-survey.html.
  5. [5]Gartner, 2024. Gartner Predicts That 90% of Finance Functions Will Deploy at Least One AI-Enabled Technology Solution by 2026. Gartner Newsroom. https://www.gartner.com/en/newsroom/press-releases/2024-09-12-gartner-predicts-that-90-percent-of-finance-functions-will-deploy-at-least-one-ai-enabled-tech-solution-by-2026.
  6. [6]Thomson Reuters Institute, 2024. Future of Professionals Report 2024. Thomson Reuters. https://www.thomsonreuters.com/en/reports/future-of-professionals.html.
  7. [7]Magesh, Varun et al., 2024. AI on Trial: Legal Models Hallucinate in 1 of 6 (or More) Benchmarking Queries. Stanford Institute for Human-Centered AI (HAI). https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-or-more-benchmarking-queries.
  8. [8]Chui, Michael et al., 2023. The Economic Potential of Generative AI: The Next Productivity Frontier. McKinsey & Company. https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier.
  9. [9]Brynjolfsson, Erik et al., 2023. Generative AI at Work. National Bureau of Economic Research, NBER Working Paper 31161 (published in Quarterly Journal of Economics, 2025). https://www.nber.org/papers/w31161.
  10. [10]Forrester Consulting, 2024. The Total Economic Impact of Glean. Forrester Consulting (commissioned by Glean). https://tei.forrester.com/go/Glean/workAIplatform/?lang=en-us.
  11. [11]Singla, Alex et al., 2024. The State of AI in Early 2024: Gen AI Adoption Spikes and Starts to Generate Value. QuantumBlack, McKinsey & Company. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024.
  12. [12]Peng, Sida et al., 2023. The Impact of AI on Developer Productivity: Evidence from GitHub Copilot. arXiv 2302.06590. https://arxiv.org/abs/2302.06590.
  13. [13]Bain & Company, 2024. AI in Financial Services Survey Shows Productivity Gains Across the Board. Bain & Company. https://www.bain.com/insights/ai-in-financial-services-survey-shows-productivity-gains-across-the-board/.