Six Ways AI Actually Changes How a Business Competes — With Real Company Examples

The previous article asked: why do most companies use AI and see no change in their financials?

The answer: because most AI optimizes the cost side — not what Drucker called "marketing and innovation." Efficiency improves, but that efficiency never gets converted into anything measurable on the P&L.

This article goes one level deeper: what does AI that enters real business logic actually look like?

Not technical descriptions — real examples of companies already using AI to change how they charge, reshape their cost structure, or redefine their customer relationships. And more importantly: whether your company can get there.

Pattern 1: From "Selling Tools" to "Selling Outcomes"

The traditional service or software logic is simple: clients pay a monthly fee or a day rate, regardless of whether results materialize.

AI makes a different logic viable: don't sell a promise — sell completed tasks.

In September 2024, Zendesk announced the first outcome-based pricing in the CX industry: $1.50 per successfully resolved ticket, with no charge if the issue isn't resolved. That same month, Intercom's AI agent Fin launched at $0.99 per resolution, with resolution rates climbing from 27% at launch to 67%. HubSpot followed in 2026 with its Breeze Prospecting Agent — $1 per qualified lead delivered.

The market is moving fast. A Growth Unhinged survey of 240 software companies found that over the past 12 months, pure seat-based pricing dropped from 21% to 15%, while hybrid pricing climbed from 27% to 41%. An a16z survey from December 2024 found that 43% of enterprise buyers now list outcome-based pricing as an important factor in purchasing decisions.

This logic applies equally well to smaller service businesses — and AI makes it operationally feasible for the first time. Marketing agencies can shift to charging per qualified lead delivered. SEO firms can charge per keyword ranking achieved. HR consultants can charge per successful hire. For buyers, the benefit is eliminating the gamble on whether it'll work — you pay when results arrive. For sellers, it ties company revenue directly to client outcomes. That's a fundamentally different competitive position, and a genuine signal of confidence.

Pattern 2: From "Selling Labor" to "Selling a Platform"

Accounting, legal, consulting, outsourced customer service — these fields share a hard ceiling: take on one more client, hire one more person. Growth and headcount track almost the same line.

AI creates the possibility of decoupling them.

A Karbon survey of AI-adopting accounting firms found that each employee saves an average of 18 hours of routine work per month — the equivalent of more than two additional months of effective capacity per year. Tech-forward firms generate 39% more revenue per employee than their peers. At Harvard Law School's CLP, AI tools compressed complaint drafting from 16 hours down to 3–4 minutes.

The Harvey × A&O Shearman case confirms the direction at scale: after deploying Harvey across 4,000 lawyers globally, contract review time shortened by 30% and each attorney saved 2–3 hours of routine work per week. A Gartner survey of 321 customer service leaders in December 2025 found that 55% are now serving more customers with the same headcount.

This is where gross margin structure actually changes: growth no longer fully depends on adding headcount. Taking on one more client increases compute costs — not a full salary. That's the shift that moves company valuations.

This doesn't mean people disappear. People still handle final judgment, edge cases, and relationship maintenance. What changes is the ratio between headcount and revenue — and that's exactly what investors and buyers care about.

Pattern 3: From "Customer Service Cost" to "Revenue Engine"

In most companies' books, customer service is a pure cost line. Problem comes in, gets resolved, money goes out, done.

Klarna's case shows what a different model looks like.

Klarna's AI assistant handled 2.3 million conversations in its first month — equivalent to the workload of 700 full-time agents. Customer wait time dropped from 11 minutes to 2 minutes. Cost per transaction fell from $0.32 to $0.19. The estimated profit contribution in 2024: $40 million.

But the more important part: Klarna's AI assistant simultaneously offers shopping recommendations — price comparisons, product selection, personalized suggestions — directly shortening the path from "asking a question" to "placing an order." Customer service shifted from cost center to sales channel.

This logic holds at much smaller scale too. Carrefour Taiwan launched an AI Sommelier, letting shoppers get wine recommendations through an AI conversation. Within two months: 30,000 users, and 70% of users who received a recommendation actually purchased. Not more efficient customer service — the service interface itself drove sales. Shopify data shows that merchants using AI personalized recommendations see 20–30% conversion rate improvement.

Worth being honest about: Klarna partially reintroduced human agents in 2025 because quality on complex issues dropped. AI customer service has real boundaries. Its role should be to expand human capability, not replace human judgment. The same entry point can lead to different outcomes — but that requires redefining what "success" means for customer service: resolution rate, or conversion rate? Both are worth tracking, but you need to be clear on which one first.

Pattern 4: From "Low-Frequency Product" to "High-Frequency Service"

Many industries run on low-frequency transactions. Insurance: one policy per year, interaction only when something goes wrong.

Progressive's Snapshot program changed that logic. By continuously tracking driving behavior through an app, safe drivers earn discounts while higher-risk drivers face higher premiums. Progressive transformed from "an insurance company you hear from once a year" into "a daily driving risk management service." Snapshot has now distributed over $1.2 billion in discounts to safe drivers.

Lemonade went further. 96% of first-time claims require no human intervention; 55% are processed fully automatically in seconds — versus the industry standard of weeks. In-force premium reached $1.24 billion in 2025, up 31% year-over-year. The trust built through fast claims handling drives customers to proactively cross-purchase other products — from renters insurance into auto, then pet insurance.

This logic isn't limited to large insurers. Gyms that implemented AI personalized interaction saw 25% better member retention and 12–18% higher annual spend. Financial advisors using AI doubled their daily client touchpoints, with retention improving 2–5%. Clinics, repair services, tutoring centers — any business that's traditionally low-touch: if AI enables sustained customer interaction, both the depth of the relationship and the ceiling on pricing change.

The shift in contact frequency doesn't just improve experience. It changes the pricing logic and the nature of the customer relationship itself.

Pattern 5: From "Doing Projects" to "Selling a Platform"

The traditional consulting and research firm model: charge by people and days. Run a project, send a team, close it out a few weeks later.

McKinsey packaged its procurement consulting methodology into Spendscape — a SaaS platform where clients subscribe and directly query supplier analysis and procurement data in real time. Instead of commissioning McKinsey to run a project each time, clients subscribe to the platform. The same core knowledge, shifted from "sending people every time" to "one version serving many clients simultaneously."

BCG's numbers are even more direct: AI service revenue at $2.7 billion per year, representing roughly 20% of total consulting revenue. The strategic direction: converting billable hours into a hybrid model of proprietary software subscriptions and outcome-based pricing.

This isn't the exclusive domain of large consulting firms. Stefan Debois spent 15 years as a consultant before turning his assessment methodology into Pointerpro — a SaaS platform where clients fill out questionnaires and automatically receive personalized reports. Today: $287,000 MRR, 30%+ annual growth, customers in 65 countries, including Deloitte and Adobe. From billing by the day to one platform serving clients worldwide — the business logic is the same thing.

The hard part of this transition isn't the technology. It's whether the company is willing to structure its core knowledge, and how to price it so clients see a subscription as better value than hiring consultants. These are strategic questions, not tool selection questions.

Pattern 6: From "Standard Pricing" to "Personalized Pricing"

The traditional model: everyone sees roughly the same products, the same discounts, the same recommendations.

Amazon makes 2.5 million price adjustments per day. After Stitch Fix deployed AI personalization, revenue per active client grew for six consecutive quarters — the latest quarter hit $559 per person, up 5.3% year-over-year. Netflix's recommendation engine saves the company more than $1 billion per year in subscription churn: 80% of content watched comes from AI recommendations, not active search.

These three companies sell something different to every person and collect different amounts from each — but they're running the same underlying system. That's the compounding edge of personalized pricing: marginal cost approaches zero, but the value proposition is different for each customer.

You don't need to be at their scale to start this logic. OneClickUpsell implemented AI dynamic offers and went from $6,000 in upsell revenue in month one to $41,000 by month three — 160% monthly growth. Envive research found that personalized recommendations drive 31% of revenue in sessions where users clicked on recommendations; 89% of marketers report positive personalization ROI.

If AI moves conversion rate from 3% to 4%, or lifts average order value by 10%, those numbers go directly into revenue. Not saving time — changing how you sell.

Not Every Company Can Get There

The six patterns above aren't for every company to chase. The question is: which type is yours?

Three categories make this clearer.

A-class: AI likely to fundamentally change the business. Information-dense, software-heavy, data-rich — cloud platforms, fintech, ad platforms, cybersecurity, digital-native e-commerce. These companies' core assets are data and algorithms; AI directly upgrades their core competitive capability. For them, the six patterns are a prioritization question, not a feasibility question.

B-class: The biggest group — AI can improve efficiency, but whether it transforms depends on execution. Banks, insurance, traditional software, retail, industrial manufacturing, logistics, enterprise services. Same industry, good execution widens the gap; bad execution just adds another tool subscription fee. This group's AI strategy determines their competitive position five years from now.

C-class: AI hardest to fundamentally change. This is the least-discussed category, but the most important one to understand clearly.

Type	Representative Companies	Why AI Can't Fundamentally Change It
Regulated Utilities	Duke Energy, ConEd	Revenue set by regulated asset base and rate reviews; AI can do predictive maintenance but won't turn the grid from heavy-asset to light-asset
Oil & Gas Pipelines	Kinder Morgan, Enterprise Products	Business model is long-term contracts, pipeline capacity, and tariffs; AI can optimize maintenance but can't change the charging logic
Commodity Producers	Exxon, BHP, Rio Tinto	Profits driven by oil, copper, and geopolitical prices; AI helps exploration and maintenance but commodity prices remain the dominant variable
Airlines	Delta, United	Core costs: fuel, aircraft, maintenance, pilots, airport slots; product experience constrained by physical seats and routes
Rail / Freight Infrastructure	Union Pacific, CSX	Delivery speed constrained by track, hubs, and physical network; AI can improve scheduling but can't accelerate physical delivery
Cement, Steel, Chemicals	Nucor, Dow	Costs mainly energy, raw materials, commodity cycles; AI can optimize processes but can't change the commodity cycle itself
Low-Margin Physical Retail	Kroger, Dollar General	AI can improve inventory and pricing, but thin margins, heavy logistics, and intense price competition mean efficiency gains get eaten by competition

McKinsey's analysis puts it directly: "In commodity products industries, AI impact mainly comes from production process optimization, not business model restructuring." Utilities can use AI for demand forecasting and equipment maintenance, but shareholder returns are still determined by regulated rate-of-return and rate reviews.

These companies aren't failing to use AI — many already are. But AI is more like: making the existing machine run smoother, not replacing the machine itself.

One exception worth noting: industry label does not equal destiny. Walmart isn't just a retailer — it has a massive advertising business and supply chain data at scale. JPMorgan holds the most complete financial transaction database in the world. GE Vernova and Schneider Electric are direct beneficiaries of AI data center buildout. The real question is whether this company has data and process advantages that AI can amplify — not just what sector label it carries.

Knowing Your Category Matters More Than Chasing Patterns

Choosing among the six patterns isn't the hardest question. The hardest question is: which path actually fits your company?

For C-class companies, AI's focus should be operational efficiency — lower per-service cost, better maintenance prediction, less administrative friction. Chasing business model rewrites is usually waste.

For B-class companies, at least one of the six patterns has a natural entry point worth thinking hard about. Not all of them apply, but there's usually one that fits your cost structure and customer relationship as they actually exist today. Zendesk's outcome pricing, Klarna's service-to-revenue shift, McKinsey's platformization — none of these started with tool selection. They started with "which part of my business logic can be redesigned."

For A-class companies, the question isn't whether to do it, but execution speed and which pattern goes first.

Tools live at the execution layer. Once direction is clear, picking tools is fast. Without clear direction first, the usual result: tools deployed, problem still there.

AI's real competitive value isn't making employees busier or faster — it's making the company's revenue model, cost model, or customer relationship model become something structurally different.

Want to figure out which category your company is in?

Not selling software, not pushing tools. Just talking through your business logic to see if there's something worth thinking harder about.

Talk to Barry

References

Back to Blog