Start Here: AI Battlegrounds in Plain English

A five-minute guide for anyone who wants to follow the AI fights in the news. No technical background needed.

What's actually new

For decades, software has been a tool. You click, and it does what you told it. The newest AI is different: it can do work on its own. It can answer a customer, write code, draft a doctor's notes or fill out a purchase order. When software can do work that people used to do, a lot of money changes hands. That's why companies are fighting over it.

The seven layers, explained with a restaurant

Every AI product sits on a stack of seven layers. Think of a restaurant.

LayerIn a restaurant
7 Business outcomesThe diner pays, is happy and comes back
6 Workflow and agentsThe kitchen line and waitstaff getting orders out
5 Domain platformsThe type of restaurant and its franchise rules
4 ModelsThe chef's skill
3 DataThe pantry and the recipe book
2 ComputeThe kitchen equipment
1 EnergyPower and water to the building

Every battleground on this site is a fight over one or more of these layers: who owns the kitchen, who has the best chef, and who gets paid when the diner comes back.

Incumbents and challengers

Incumbents are the established companies that already lead a market. They have customers, revenue and a lot to lose. Challengers are newer companies, or companies from other markets, using AI to take that business away. Often they compete and cooperate at the same time: the same two companies can be partners in one deal and rivals in the next.

How to read a battleground card

  • Battleground 10 · Layers 6 + 7

    Who Does the Customer-Service Work?

    Who, or what, does the work, and who answers for it

    The management question

    Are you designing work, or still counting heads?

    • Incumbents: Teleperformance, Salesforce
    • Challengers: Sierra, Decagon, Parloa
    Explore battleground
  1. Number and layers — where the fight sits in the stack. Layers 6 and 7: who does the work, and who gets paid for the result.
  2. The question — what's being fought over.
  3. The shared contest — the prize everyone wants.
  4. The management question — what a business leader should ask about their own company.
  5. Incumbents and challengers — the main players on each side.

Cards marked Emerging are fights still taking shape.

Glossary

AI agent
AI software that takes actions and completes tasks, not just answers questions.
Ambient AI
AI that listens in the background, for example during a doctor's visit, and writes up the notes automatically.
Challenger
a newer company, or one from another market, trying to take business from the established players.
Compute
the computing power, mostly specialized chips in large data centers, that AI needs to run.
Context layer
the organized business information an AI needs to act correctly, such as what counts as a "customer."
Data center (AI factory)
a large building full of computers that trains and runs AI. It uses a lot of power and water.
Domain platform
the main software system for a business area or industry, such as the health-record system in a hospital.
Ecosystem
the network of companies (suppliers, partners and rivals) that work around a major platform.
Forward-deployed engineer (FDE)
an engineer from a technology company who works on site with a customer to get AI working in its business.
Global systems integrator (GSI)
a large consulting firm, such as Accenture or Deloitte, that implements technology for big companies.
Hyperscaler
one of the giant cloud providers: Amazon Web Services, Microsoft Azure and Google Cloud.
Incumbent
an established company that already leads a market.
Model
the "brain" of an AI system, such as the technology behind ChatGPT, Claude or Gemini.
Moratorium
a temporary ban; on this site, usually on building new data centers.
Neocloud
a newer cloud company built specifically to rent out AI computing power, such as CoreWeave.
Ontology
a shared map of what a company's data means and how it connects.
Outcome pricing
paying for results, such as each customer problem solved, instead of paying per user.
Procure-to-pay
the business process from requesting a purchase to paying the supplier.
Seat-based pricing
paying a fixed fee for each person who uses the software.
Vibe coding
building software by describing what you want in plain language and letting AI write the code.
Workflow
the sequence of steps people and systems follow to get a piece of work done.