THE SOFTWARE BUDGET FOLLOWS THE WORK
The center
of gravity
is moving.
Claude’s native documents and spreadsheets reveal a much larger change: frontier AI providers are becoming the place where work lives.
The customer relationship follows.
The important part of Claude Docs is where the document stays.
For years, the familiar AI workflow ended with a small act of departure. Ask a question. Get a draft. Copy it into Word or Google Docs. Move the numbers into Excel. Put the design in Canva. Take the discussion back to Slack. Update the customer record somewhere else.
The AI helped with the work. The applications still housed it. Every departure renewed the importance of another subscription.
Claude’s own documents and spreadsheets change that arrangement. The draft can become a living document inside Claude. The budget can become a shared sheet. Other people can join the work there. You have fewer reasons to leave.
My prediction is that frontier AI providers will become the center of gravity for most people’s digital work. As they provide office productivity, team collaboration, project management, and customer management natively, they will capture more wallet share from the applications we currently pay for separately.
The frontier model is the engine. The place built around it becomes the business.
The document
is a place.
Anthropic’s current documentation describes Claude Docs as rich-text documents saved in Claude, with simultaneous editing, comments, and access controls for collaborators. Claude Sheets puts people and Claude in the same spreadsheet, working with cells and formulas. Sheets remains in beta. These are native working surfaces, distinct from asking Claude to operate an external file. [1, 2]
On November 10, Anthropic launched Sheets and Whiteboards in beta and made Claude Data generally available. Its explanation was unusually revealing: work previously moved into other programs; now documents, boards, sheets, and dashboards can remain in Claude for teams to edit together. [3]
That is a change in destination.
A generated file is useful. A shared, persistent place is more consequential. It accumulates the discussion, revisions, decisions, and people around the file. Tomorrow’s task begins closer to yesterday’s understanding.
The change becomes visible when a colleague asks, “Where is the launch plan?” and the answer is a Claude link. At that moment, Claude is participating in the team’s daily coordination. It has become somewhere people go together.
Follow the
next renewal.
Imagine a small consulting firm preparing a client launch next year. The partner works through the offer with her AI. A colleague joins the proposal document. The budget is beside it. The schedule has owners and dates. The system retains the client’s approved requirements and the team’s decisions within the permissions they have granted.
When the launch changes, the team asks for the consequences. Which promises are affected? What happens to the budget? Which client conversation needs to happen first? The value comes from carrying the meaning of the work across its different forms.
At renewal, the partner looks at the software bill. The design seat used for occasional presentations has become less necessary. The separate project board duplicates a working plan the team already shares. A lightweight customer tracker can live beside the proposal and follow-up history.
One category at a time, a separate purchase becomes redundant.
The frontier provider wants that spending. It can include another useful capability because the addition makes the whole relationship more valuable and harder to replace. It can earn more from a person even while that person spends less across the old collection of vendors.
Neither outcome is automatic. Intelligence consumes compute. Collaboration requires storage, access management, recovery, and reliable service. Heavy use can carry usage charges. The provider has to deliver the work at an economic cost. But it is competing for an existing budget spread across many invoices, which gives it room to build a broader offer.
The customer does not need to learn the economics. She needs to reach the next renewal and realize she has stopped using something she still pays for.
A SMALL EXPERIMENT IN SOFTWARE ECONOMICS
Where does
the work live?
Follow one client launch through three arrangements. Select a stage to see what changes.
A conceptual progression, not a claim that Claude currently supplies a complete native CRM or project management system.
Collaboration
turns the key.
A personal AI habit becomes much more powerful when other people can participate without abandoning the context already developed.
One person begins with a brief. A second adds the sales constraints. A third corrects the budget. Their shared workspace becomes useful because each can see and change the actual work. The AI can help with what they are building together.
That is the bridge from an individual subscription to a team’s software spending. Invitations carry the working environment outward. The colleague arrives to review a proposal and discovers she can build her next one there too.
This is why native collaboration matters more than another impressive demonstration of a model writing a document. A model can generate a beautiful draft without changing anyone’s software budget. A team that keeps returning to the same shared work creates a reason to consolidate that budget.
Over time, the expectation reverses. We will ask our primary intelligence provider why a familiar capability is missing. Documents, sheets, a project plan, a shared discussion, and a customer history will feel like parts of the service we already bought.
The standalone app will increasingly have to explain the additional value of another paid relationship.
A connector
is the bridge.
There are two commercial stages here, and confusing them makes the forecast weaker.
Anthropic’s interactive connectors can bring partner applications into Claude. Its November 10 announcement includes a Salesforce plugin in beta, along with other business tools. Canva describes its own Claude Design partnership as a route from an AI draft into Canva’s editable design environment. Those integrations can strengthen the existing supplier even when Claude becomes the starting point. [4, 5]
Opening Salesforce through Claude does not cancel the Salesforce contract. Opening Canva inside an AI workflow does not make Canva’s design infrastructure free.
First, the interface relationship moves. Later, native capability can move the spending.
CRM makes the distinction clear. A shared table of customers is easy to imagine. A dependable CRM must also preserve relationship history, resolve duplicate records, manage permissions, maintain commitments, connect communications, and recover when an action fails. A business has to trust that it will still work next Tuesday.
Frontier providers will build more of that operating layer. For a new business with modest requirements, it will be easier to begin there than to buy a separate CRM and connect it afterward. For an enterprise with years of records, complex approvals, and demanding controls, the existing system will remain valuable much longer.
The fastest displacement will come from work that never enters the old application in the first place. Each new firm, new project, and new team gets to choose where its history begins.
THE INCUMBENTS SEE THE SAME DESTINATION
Microsoft
is moving
toward it, too.
On November 17, Microsoft announced general availability of Copilot Home and Office in Copilot, including editing and coauthoring. Workspaces was available in Frontier, and Autopilot was beginning its Frontier rollout. Microsoft is bringing the office into the AI environment while drawing on models from OpenAI and Anthropic. [6]
This is the incumbent’s strongest response: become the intelligence relationship through which the customer does the rest of the work.
Microsoft can retain wallet share where it succeeds at that. Canva can earn its place through design capabilities people continue to value. Salesforce can remain a consequential system of record. The pressure is on every provider to justify its role inside the new arrangement.
The frontier labs have an opening because many people already begin with them. The incumbents have an opening because much of the world’s working context already resides in their systems.
The contest is over who becomes the place you stay.
The budget
changes shape.
I expect most people to pay their primary AI provider for an expanding range of everyday capabilities. Many will stop paying separately for the ordinary functions they once bought through Canva, Microsoft productivity seats, Slack, project management applications, and lightweight CRM products.
Specialists will keep paying for specialized value. Large organizations will keep consequential systems where migration costs and operating requirements justify them. But the frontier provider will absorb a growing portion of the routine work that once supported a separate seat.
That changes what software companies sell. Some will continue to own the customer relationship. Others will supply capabilities underneath someone else’s broader service. A specialist can prosper as an input, but its economics will depend on the value it adds to the whole offer and the bargaining power it retains.
It also changes the frontier provider’s priorities. A slightly better answer helps win a comparison. A working customer history, shared budget, and dependable follow-up process help win a continuing relationship. The incentive expands from improving the model to maintaining the environment around a person’s work.
This is the wallet-share mechanism I described in The Last Subscription, now becoming visible in the architecture of the product. Useful integration gives spending a new destination.
As that environment becomes more central, the ability to leave becomes more important. Exportable files are a beginning. Practical exit also requires usable records, permissions, history, and a working handover. The person’s own methods and memory must remain distinguishable from an employer’s confidential information.
A primary workspace will be powerful because it understands the work. It earns that position by continuing to serve the human whose work it is.
Sources & further reading
Product capabilities checked November 18, 2026. The wallet-share argument and future scenes are my forecast.
- Anthropic · Get started with Claude Docs. Native documents, shared editing, ownership, and permissions.
- Anthropic · Get started with Claude Sheets. Native shared spreadsheets. The beta currently lacks pivot tables, protected ranges, Excel macros, and live feeds from connected apps; complex spreadsheet work still has reasons to remain elsewhere.
- Anthropic · Introducing Claude Whiteboards, Claude Sheets, and Claude Data, November 10, 2026.
- Anthropic · More ways to work with Claude: Sheets, Whiteboards, and Interactive Connectors, November 10, 2026.
- Canva · Introducing Canva in Claude Design by Anthropic Labs. A concrete example of the existing application retaining a role in an AI workflow.
- Microsoft · Copilot at Microsoft Ignite: Expanding what’s possible with intelligence built for work, November 17, 2026.
Continue: The Last Subscription · The First Stack Generation