See also (wiki): ai-era-business-models · ai-maturity-models · workflow-redesign
Executive Summary
- MIT CISR (Peter Weill, Stephanie Woerner, Ina Sebastian, Gayan Benedict) published “Business Models in the AI Era” on October 16, 2025, drawing on a longitudinal dataset of 2,378 companies surveyed across 2013, 2019, 2022, and 2025. This is the most data-backed framework in circulation for C-suite discussions about AI’s business-model implications.
- The prior-generation map has already collapsed. Between 2013 and 2025, the Supplier model fell from 46% of companies to 15%. Omnichannel fell from 24% to 4%. Ecosystem Driver rose from 12% to 58%. Ecosystem participation broadly rose from 30% to 81% of companies.
- Financially, only one of the four original digital models — Ecosystem Driver — delivers above-industry-average revenue growth in 2025, and does so by roughly 6 percentage points.
- The new AI-era framework maps businesses on two axes: action (Assist vs Represent the customer) and execution (Structured vs Adaptive process). The four resulting models are Existing+, Customer Proxy, Modular Curator, and Orchestrator. Most companies will operate in several simultaneously.
- The practical use of this framework is internal alignment, not external disruption. It gives a CEO, CSO, and CFO a shared vocabulary for deciding which customer outcomes AI should assist with today, which it should represent inside guardrails, and which require redesigning the underlying process entirely.
Where the Digital Business Model Map Stands in 2025
MIT CISR has tracked business model evolution against consistent survey questions for twelve years. The 2,378-company dataset shows a decisive shift in where companies actually sit:
| Model | 2013 | 2025 | Shift |
|---|---|---|---|
| Supplier (sells through others) | 46% | 15% | −31 pp |
| Omnichannel (digital + physical channels) | 24% | 4% | −20 pp |
| Modular Producer (plug-and-play service) | 18% | 23% | +5 pp |
| Ecosystem Driver (destination connecting customers and providers) | 12% | 58% | +46 pp |
Ecosystem participation overall — companies either leading or participating in digital ecosystems — grew from 30% to 81%. Ecosystem Driver is the only one of the four that currently outperforms its industry average on revenue growth, by approximately 6 percentage points. For a Fortune 500 board, this is the decade-long reason digital strategy kept returning to “platform” and “ecosystem” language: the models that treat the customer as the endpoint of a bundle of third-party services grew, while those that treated the customer as the endpoint of a single product line eroded.
The AI-era framework rests on this foundation. Agentic AI does not replace the direction of travel; it accelerates it and makes two additional business-model moves possible.
The AI-Era Framework: Two Axes
MIT CISR shifted the framework’s axes to reflect what AI changes:
Vertical — Action on behalf of the customer.
- Assist: AI helps a customer reach an outcome, but the customer stays in the driver’s seat.
- Represent: AI autonomously achieves the outcome within guardrails the company and customer set.
Horizontal — How the work gets executed.
- Structured: the company specifies the process; AI runs predefined flows; humans in the loop review/approve, humans at the helm set goals, incentives, and constraints.
- Adaptive: the company specifies the outcome; AI assembles the process itself; humans at the helm monitor rather than direct.
The four quadrants:
| Model | Axis position | Plain-English description | Illustrative financial-services example |
|---|---|---|---|
| Existing+ | Assist + Structured | AI strengthens today’s processes | Advisor uses AI to analyze a client’s holdings and generate personalized recommendations |
| Customer Proxy | Represent + Structured | AI executes on customer’s behalf inside fixed rules | Portfolio auto-rebalances within preset parameters |
| Modular Curator | Assist + Adaptive | AI assembles a custom bundle from modules (including third-party) | Firm builds one-off client packages combining investments, insurance, credit from multiple providers |
| Orchestrator | Represent + Adaptive | AI autonomously delivers customer outcomes across an ecosystem | Fully managed wealth solution continuously optimized without customer input |
The concrete Orchestrator example MIT CISR cites from outside financial services is Amazon’s “Buy for Me” feature, which requests purchases from other brands’ sites for products Amazon does not sell directly.
What One New Zealand Tells You About Sequencing
MIT CISR’s anchor case is One New Zealand, a telco making an explicit pivot to AI-driven operations. The sequencing is instructive for any mid-market executive sketching a three-year plan.
- 2024: 15 AI use cases cleared the internal margin hurdle.
- 2025 target: 50 AI solutions deployed; 30 launched or underway by publication.
The mix is not evenly distributed across the four models. It is weighted toward Existing+ today, with Customer Proxy in controlled rollout, Modular Curator in early operational deployments, and Orchestrator reserved as the future state.
- Existing+: Knowledge agents resolve 60% of consumer prepaid queries and 40% of enterprise queries in beta. Marketing copywriting agents cut audience segmentation time 60%.
- Customer Proxy: Service agents handle upgrades, disconnections, and support tickets with humans retained in the loop.
- Modular Curator: Task-based agents during weather events verify power and cell status, forecast demand, and recommend generator deployment in minutes rather than hours.
- Orchestrator: Network optimization and fully autonomous marketing campaign creation are specified as the future state, not the current state.
The pattern: structured before adaptive, assist before represent. The company crossed the Customer Proxy threshold only after several quarters of Existing+ deployment produced the telemetry, guardrails, and human oversight patterns that made “represent the customer” a defensible choice.
Source Credibility
HIGH. MIT CISR is the single most-cited institutional research program in the corpus. The Weill/Woerner team has maintained the same instrument across twelve years. The 2,378-company sample is larger than BCG AI Radar 2026 (CEO sample) and comparable in rigor to McKinsey’s State of AI longitudinal. This is a Tier 1 source (October 2025 publication).
Caveats worth stating: the four AI-era categories are a conceptual framework tested in senior executive sessions, not a survey-measured distribution with financial outcomes per quadrant. Unlike the original digital business model data, MIT CISR has not yet published revenue growth or profit figures by AI-era category. Expect that data in the next research cycle.
Key Data Points
| Data point | Value | Source | Date |
|---|---|---|---|
| Total longitudinal sample | 2,378 companies | MIT CISR | 2013–2025 |
| Ecosystem Driver share, 2013 → 2025 | 12% → 58% (+46 pp) | MIT CISR | Oct 2025 |
| Supplier share, 2013 → 2025 | 46% → 15% (−31 pp) | MIT CISR | Oct 2025 |
| Omnichannel share, 2013 → 2025 | 24% → 4% (−20 pp) | MIT CISR | Oct 2025 |
| Ecosystem participation, 2013 → 2025 | 30% → 81% | MIT CISR | Oct 2025 |
| Ecosystem Driver revenue growth premium | +6 pp vs industry average | MIT CISR | Oct 2025 |
| One NZ consumer prepaid query resolution (Existing+) | 60% | MIT CISR case | Oct 2025 |
| One NZ audience segmentation speedup | 60% faster | MIT CISR case | Oct 2025 |
| One NZ AI solutions deployed | 30 of 50 target, end-2025 | MIT CISR case | Oct 2025 |
What This Means for Your Organization
If you lead a 200- to 5,000-employee company, the immediate use of this framework is not to declare “we are an Orchestrator now.” It is to force an honest conversation about where you already are, where your competitors already are, and which customer outcomes you are still delivering through a Supplier or Omnichannel posture that the market has largely moved past. The 46-point rise of Ecosystem Driver is not a prediction. It is a description of the last twelve years.
The more actionable question is sequencing. One New Zealand did not attempt the Orchestrator leap. They stacked Existing+ deployments until the guardrails and human-in-the-loop patterns made Customer Proxy safe, then let Modular Curator emerge in an operational niche (weather-event response) where the cost of adaptive execution was low and the upside was obvious. Every mid-market AI roadmap should copy that sequencing logic — Assist before Represent, Structured before Adaptive — and be willing to sit in Existing+ for two to four quarters longer than feels ambitious. The companies that skipped that step in prior technology cycles (cloud, mobile, data platforms) are the ones still paying down architectural debt.
The third question is which quadrants your customers will tolerate from you. A private bank’s high-net-worth clients may welcome Modular Curator but resist Orchestrator. A small-business insurance customer may prefer Customer Proxy to the manual advisory process you run today. The framework’s value is forcing you to answer this per customer segment rather than per product line.
If this raised questions specific to your organization’s positioning across the four models, I’d welcome the conversation — brandon@brandonsneider.com.
Sources
- Peter Weill, Ina M. Sebastian, Stephanie L. Woerner, Gayan Benedict. “Business Models in the AI Era.” MIT CISR Research Briefing Vol. XXV, No. 10, October 16, 2025. https://cisr.mit.edu/publication/2025_1001_BizModelsAIEra_WeillSebastianWoernerBenedict — Credibility: HIGH. Longitudinal, n=2,378, consistent instrument 2013–2025.
- Peter Weill, Stephanie Woerner. “AI-Era Business Models Framework.” MIT CISR. https://cisr.mit.edu/publication/AIEraBusinessModels_Framework — Credibility: HIGH. Companion framework page.
- Sara Brown. “How digital business models are evolving in the age of agentic AI.” MIT Sloan Ideas Made to Matter, 2025. https://mitsloan.mit.edu/ideas-made-to-matter/how-digital-business-models-are-evolving-age-agentic-ai — Credibility: HIGH. Secondary summary with author-confirmed quotes.
- MIT CISR. “What’s Your AI-Enabled Business Model?” (research-in-progress page.) https://cisr.mit.edu/content/whats-your-ai-enabled-business-model — Credibility: MEDIUM. Announcement page, limited data.
Brandon Sneider | brandon@brandonsneider.com April 2026