Large enterprises don't deploy AI to "look modern" — they deploy it because it lowers cost per service unit, and that moves margin directly on the P&L. They target six specific areas, always the same ones: customer service, back-office, supply chain, sales and marketing, finance and human resources. The logic works identically for an SMB; the only thing that changes is the price tag of the tool.
There's a lot of confusion about why Klarna, JP Morgan, Unilever or BBVA are investing hundreds of millions in AI. The short answer isn't "innovation". It's margin.
Any AI project that survives a finance committee at a corporation has to be justified in a single line: how much it lowers cost per transaction and how much it raises operating margin. If it doesn't fit in that sentence, it isn't approved. The rest — the press release, the innovation event, the photo with the CEO — is decoration.
In this article we walk through the six specific areas where large corporations are deploying AI in 2026, which KPI they move in each one and, more importantly, what a smaller company can learn from that same logic.
The premise: AI isn't automation, it's a margin lever
Classic automation lowers the cost of a specific process. A margin lever lowers cost per service unit at scale, without sacrificing quality or breaking the operating model. The difference matters.
Concrete example: a company can automate sending welcome emails (that's automation, saves time). Or it can deploy an AI agent that qualifies the lead, answers their questions, books a demo and only escalates high-intent cases to the sales team — that's a margin lever, because it lowers customer acquisition cost (CAC) and frees commercial capacity for cases where a human is actually needed.
This is how CFOs look at it:
| Concept | Classic automation | AI as margin lever |
|---|---|---|
| Metric moved | Time per task | Cost per service unit |
| Scale | Linear (more volume = nearly proportional more cost) | Sub-linear (more volume ≠ more cost) |
| P&L impact | One-off saving | Structural margin improvement |
| Who approves it | Department head | CFO or investment committee |
Through that lens it becomes clear why six areas consistently receive the first wave of investment. They're the ones with high recurring cost + repetitive processes + digitisable data. Let's go through them one by one.
The 6 areas where AI investment concentrates
Customer service and support
It's the first area on almost every corporate roadmap. The reason is mathematical: cost per ticket in a European call centre runs between €4 and €12. If a well-trained conversational AI resolves 60–70% of queries without escalation, cost per ticket drops to cents. Multiplied across millions of interactions a year, that moves operating margin several points.
Klarna (Swedish fintech, 150 million users): its AI assistant handles the equivalent workload of 700 full-time agents. The company estimates an impact of ~$40 million per year in operating margin, with resolution times dropping from 11 minutes to under 2. Figures published by Klarna itself in 2024.
Operations and back-office
Back-office concentrates highly repetitive work: processing invoices, reviewing contracts, reconciling accounts, validating documents. Historically it was outsourced to shared service centres. With generative AI, a significant share now runs without a human operator — the human only reviews exceptions.
JP Morgan with its internal system COIN (Contract Intelligence): analyses commercial loan contracts that previously required 360,000 legal hours per year. The system does it in seconds. It didn't replace the legal team — it reassigned them to higher-value work. That's the point of the case, not headcount reduction.
Supply chain and inventory
This area is less visible but delivers the largest monetary impact at industrial or retail companies. AI improves demand forecasting, optimises routes, fine-tunes stock levels per SKU and per store. Every percentage point of improvement in forecast accuracy frees weeks of immobilised inventory, releasing working capital directly.
Walmart and DHL both use AI for route optimisation and demand forecasting. Walmart has reported double-digit improvements in in-store availability and reductions in safety stock. In European logistics, DHL attributes 10–15% reductions in kilometres driven per delivery to AI-powered routing.
Sales and marketing
Here AI attacks two fronts at once: lowering acquisition cost (sharper segmentation, content generation at scale, lead scoring) and raising customer value (personalisation, recommendations, retention). In B2B, AI agents qualify leads 24/7 and only forward genuinely high-intent contacts to the sales team.
Vodafone, with its assistant TOBi, handles millions of sales and service interactions per year across more than 20 countries. The company reports resolution rates without a human agent above 70% and measurable NPS improvements in the markets where it's deployed.
Finance and reporting
Monthly close and regulatory reporting consume a surprising amount of hours at large corporations. AI accelerates reconciliations, flags anomalies before they reach the auditor and generates report narratives automatically. The most-watched KPI here is days to close: moving from 8 days to 3 days frees finance capacity for actual analysis.
The Finance Transformation practices at the Big Four (Deloitte, EY, PwC, KPMG) consistently report 30–50% reductions in close hours on projects with IBEX 35 and Fortune 500 clients, combining RPA with generative AI layers for narrative and validation.
Human resources and talent
AI in HR addresses two very specific costs: hiring (CV screening, initial interviews, scheduling) and not retaining (analysing early signals of attrition, personalising development plans). At companies with thousands of hires per year, cutting time-to-hire by two weeks has a measurable operational impact.
Unilever has published that its AI-supported selection process — video interviews analysed by AI plus cognitive games — reduced time-to-hire by 75% and allowed for more equitable candidate review. It's one of the most thoroughly documented cases in the corporate literature.
The common pattern: cost per unit falls, margin rises
If you look at the six areas together, a clear pattern emerges. AI isn't replacing people in these public cases — it's reallocating human work to where it actually adds value, while at the same time bringing cost per interaction down to a fraction.
"The question in the committee is no longer 'should we use AI?'. It's 'which KPI do we want to move this quarter, and which of our AI pilots has a clear business case to move it?'."
— CFO at a Spanish industrial multinational, quoted at a 2026 sector forum
What an SMB can copy from this strategy
Here's the interesting part. The logic used by these corporations works just as well at SMB scale — only the price tag of the tool changes.
Where a corporation invests €2 million in an AI customer service project, a 20-person company addresses the same lever with a SaaS platform at €99–300/month. The KPI is the same (cost per interaction); the percentage impact on margin can actually be larger in proportion (because personnel cost weighs more heavily on total cost at an SMB).
The four areas with the best impact/investment ratio for an SMB are, in order:
- Customer service and lead qualification. Any business receiving 30+ queries per month benefits. An AI assistant on web or WhatsApp resolves 60–70% without escalation. It's the use case where the investment pays back fastest.
- Sales follow-up. Automatic reminders, lead segmentation, draft proposals. Lifts conversion rate without hiring more people.
- Content and communication. Emails, posts, product descriptions, review responses. AI drafts the first version, a human reviews. From hours to minutes.
- Administrative back-office. Invoice processing, document data extraction, reconciliations. Less visible to the customer, but frees up hours every week.
The most expensive mistake an SMB makes is copying the technology of a large company instead of copying its logic. You don't need a custom AI system — you need to pick the KPI that hurts most and deploy a specialised tool that moves it. That's exactly what CFOs at the big corporations do, just with fatter cheques.
How a corporation decides which AI project to approve
So you understand the mindset — and can apply it in your own company — here's the real filter used in investment committees:
| Criterion | What gets evaluated | Typical threshold |
|---|---|---|
| KPI moved | Does it move a KPI the CEO looks at every month? | Mandatory |
| Payback | How many months until investment is recovered? | < 12 months |
| Scalability | Does it still work if we triple the volume? | Yes, without re-hiring |
| Regulatory risk | Does it clash with GDPR, AI Act or sector regulation? | Mitigable |
| Reversibility | Can we roll back if it doesn't work? | Yes, in < 30 days |
When an SMB applies this same filter — even informally, in a spreadsheet — its AI projects stop being experiments and become measurable investments.
Want to see which areas to target first in your company?
In 30 minutes we analyse your processes and tell you which KPI would move margin most, and with which tool — the same logic used by the big corporations, applied to your scale. Free and no commitment.
Free assessment →Quick summary
- Large enterprises use AI to improve margin, not to look innovative. Every project is approved on P&L impact.
- There are six areas that concentrate investment: customer service, back-office, supply chain, sales and marketing, finance and HR.
- The pattern is always the same: cost per service unit drops while quality holds or improves. Humans get reassigned to higher-value work.
- Public cases with figures: Klarna (~$40M/year in customer service), JP Morgan (360,000 hours/year in legal), Unilever (-75% time-to-hire), Vodafone (>70% no-agent resolution).
- An SMB can apply the same logic with SaaS platforms at €49–300/month. What matters is the KPI being moved, not the price tag of the tool.