Applications, Agents & AI Software
Layer 07 of seven · Applications, Agents & AI Software
- Why it matters
- This is the layer that turns capability into revenue. If enterprise applications cannot show a return, the spending underneath it eventually has to slow.
- The bottleneck
- Trust, integration and change management. The technical work is often the smaller half of a deployment.
- Who captures value?
- Vendors already inside the workflow, holding proprietary data and a distribution channel. Thin wrappers around a public API tend not to keep their margin.
- What could change?
- Agents that reduce headcount also reduce seat counts, which is awkward for anyone charging per user. Pricing models in this layer are still unsettled.
Companies that use AI to deliver products, workflows and services people actually buy. This is where model capability meets a budget holder with a problem.
A foundation model is a capability, not a product. Someone has to point it at a specific problem, wrap it in permissions and audit trails, integrate it with the systems where work already happens, and persuade a buyer to change a process. That work is unglamorous and it is where a great deal of the durable economics sits, because customers rarely rip out something that is wired into how they operate.
Tap or hover a box to see what it does
Enterprise platforms and agents
Software that runs institutional processes, with AI increasingly embedded as agents that take actions rather than only answer questions. Buying criteria here are auditability and integration long before they are model benchmarks.
Palantir
Sells data integration and decision platforms to governments and large enterprises, increasingly packaged around AI driven workflows.
It solved the boring problem first. Getting messy institutional data into one usable model is most of the work, and it is very hard for a customer to unpick later.
Lumpy government contracting, delivery that depends on skilled people, and a valuation that leaves little room for disappointment.
ServiceNow
Provides a workflow platform for IT, HR and customer operations, with AI agents embedded in those workflows.
AI is most valuable where the work already happens. Owning the workflow means the AI arrives inside a process rather than beside it.
Premium AI pricing has to survive procurement scrutiny, and platform consolidation cuts both ways.
Salesforce
Sells customer relationship management software with an agent platform layered over its data and workflow estate.
It holds a great deal of customer facing operational data, which is precisely the context an agent needs to be useful rather than generic.
Seat based software faces an awkward question if agents reduce the number of seats required.
Microsoft
Runs Azure, distributes AI through Microsoft 365 and Copilot, partners commercially with OpenAI, and designs its own Maia accelerators.
Few businesses touch as many parts of the ecosystem at once. It owns infrastructure, a commercial route to frontier models, and the enterprise distribution to sell the result.
Very heavy capital spending against uncertain payback, dependence on partner model roadmaps, and buyers questioning per-seat AI pricing.
Pegasystems
Sells business process automation and decisioning software to large organisations.
Deeply embedded in complex back office processes at banks and insurers, which is unglamorous work that tends to stay put.
Slower growth than newer platforms, and modernisation cycles that customers can defer.
Creative, developer and vertical software
Tools aimed at particular kinds of work, where proximity to the task matters more than general capability. Developers were the first large professional group to adopt AI tooling at scale.
Adobe
Provides creative and document software with generative features built into its established tools.
Distribution into professional creative workflows, plus a commercially licensed training data position that matters to enterprise buyers.
Generative tools have lowered the barrier for competitors, and pricing for creative software is under more pressure than it used to be.
GitHub (Microsoft)
Hosts code and ships Copilot, the coding assistant most developers encountered first.
Developers are the fastest adopting user group in the entire ecosystem, and the tool that sits in the editor has an unusually strong habit advantage.
Coding assistants are now a crowded category with capable rivals and rapid switching.
ImmunityBio
A biotechnology company applying computational methods within immunotherapy development.
An example of AI as a research tool inside a specialist domain rather than as the product itself.
Clinical and regulatory risk dominates. The AI angle is secondary to trial outcomes.
Physical AI
Perception and control models running on hardware in the real world rather than in a data centre. Tesla is the most visible listed example, through driver assistance and autonomy, a planned robotaxi service, the Optimus humanoid programme and custom inference silicon in its vehicles.
Tesla
Builds electric vehicles and develops driver assistance and autonomy software, a planned robotaxi service, the Optimus humanoid robot programme and custom inference silicon for its vehicles.
The clearest listed example of physical AI, where perception models run on custom hardware in the real world rather than in a data centre.
Autonomy timelines have repeatedly slipped, regulatory approval is jurisdiction by jurisdiction, and the core vehicle business faces price competition.
How to think about the economics
- GrowthHigh
How quickly demand in this part of the ecosystem is expanding.
- Capital intensityLow
How much money has to be spent up front before revenue arrives.
- Competitive moatHigh
How difficult it is for a credible new entrant to take the business.
- Customer concentrationLow
How much revenue depends on a small number of buyers.
- Disruption riskHigh
How exposed the layer is to a technical or commercial shift.
This is a framework for thinking about the economics of a layer, not a recommendation. An important AI company, a strategically advantaged company, an investable security and an attractively valued security are four different things.
Where value may accrue
Distribution, workflow integration, proprietary data and network effects. The moat is rarely the model, and almost always the surrounding process.
Key terms in this layer
- Agent
- Software that uses a model to take actions on your behalf, not just answer questions. It can browse, click, buy and send with real permissions.
- ExampleAn agent that books your travel end to end, not one that just suggests flights.
- Related termsFoundation modelInference
Related questions
What is agentic AI?
Most AI use is still reactive: you ask, it answers. An agent is given a goal instead, and works out the intermediate steps, calling tools, reading data and looping until it decides the task is finished. That shift is why permissions, audit trails and identity suddenly matter so much. A chatbot that is wrong wastes your time. An agent that is wrong can take an action on your behalf, which is a different category of problem.
What is the difference between training and inference?
Training is not a single event. Pre-training is the long, expensive phase where a model learns general patterns from very large amounts of data. Post-training shapes that raw model into something useful and better behaved, often using human feedback and reinforcement learning. Fine-tuning adapts an existing model to a narrower task or domain, and can be done repeatedly and comparatively cheaply. Inference is the model actually answering, which happens billions of times a day. Much of the current demand for chips, power and data centres is driven by inference at scale rather than by headline training runs alone.
What is a foundation model?
A large, general purpose model trained on a very broad dataset, which can then be adapted to particular tasks. The GPT, Claude, Gemini and Llama families are all foundation models. They are called foundations because applications, tools and agents get built on top of them. Training one is an expensive bet that sitting at the base of the stack is where lasting value accrues, which is precisely the question the rest of this site is trying to help you think about.
Sources and further reading (3)+
- 1Investor relations and product documentation · Palantir
- 2Investor relations · ServiceNow
- 3Investor relations, autonomy and Optimus updates · Tesla
The AI ecosystem changes rapidly. Company positions, technologies and market data reflect information available at the date above.