When the Model Becomes the Workflow

A few years ago, most AI demos looked the same. You asked a question, and the model answered it. Maybe it summarized a document, drafted an email, or generated a few lines of code. The intelligence was impressive, but the work still belonged to the user.

That boundary is starting to disappear.

This week, Anthropic released a set of finance-focused AI agents built for specific workflows inside financial services. The agents can generate pitchbooks, perform KYC screening, assist with credit memos, automate reporting tasks, and support month-end close processes through Claude Cowork, Claude Code, and Claude Managed Agents.

At first glance, it looks like another vertical AI release. What matters more is that model companies are starting to package workflows themselves.

For most of the AI cycle so far, model providers positioned themselves as infrastructure. They supplied intelligence while software companies built the applications on top. Over time, SaaS companies settled into a clear division of responsibility: the model company handled reasoning, while the software company owned the workflow.

Anthropic’s finance agents blur that line. These are pre-packaged systems designed around repeatable professional tasks, with the workflow already embedded into the product. Once model companies begin packaging workflows directly, many horizontal SaaS products lose part of the value they once controlled.

The important detail is not that Anthropic launched finance tools. It is that they launched workflow products.

A chatbot helps a finance team think through a task. A workflow agent performs part of the task directly, understanding the sequence, the documents involved, the expected outputs, and the operational context around the work itself.

The product is no longer limited to intelligence on demand. It now packages labor into software, which puts traditional SaaS companies in a more vulnerable position.

Enterprise software has historically been built around organizing repeatable workflows: forms, dashboards, approval systems, and reporting layers that structured human work.

As AI agents begin executing parts of that work directly, customers evaluate software differently. Interface quality and seat count matter less than measurable outcomes.

How much time did it save?

How much work did it eliminate?

How much faster did the process move?

Products that simply wrap interfaces around standardized workflows become easier to replace once those workflows themselves become programmable.

The companies moving fastest are not trying to compete directly with model providers. Instead, they are moving deeper into domain expertise.

Rather than building generic AI wrappers, they are focusing on proprietary workflows, industry-specific logic, governance layers, integrations, and operational trust. They are combining AI with services, implementation support, and workflow knowledge that templates alone cannot reproduce.

At the same time, many teams are rethinking pricing. If AI performs more of the labor, charging purely per seat becomes harder to justify. Companies are starting to price around completed work, processed transactions, or measurable outcomes, which changes product strategy as much as technology strategy.

This issue of Redeployed is brought to you by Tecla: As AI vendors move from supplying models to packaging entire workflows, the advantage no longer comes from simply adding AI features. The teams moving fastest are combining AI with domain expertise, governance, and systems integration, bringing in engineers who can translate real business workflows into reliable products. Tecla helps companies hire senior tech talent in the U.S. and nearshore who already work across AI systems, industry workflows, and production environments, so teams can build defensible products instead of generic AI wrappers.

Packaging workflows sounds simple until those workflows touch the real world.

Finance is full of ambiguity, edge cases, approvals, compliance requirements, and exceptions that do not fit neatly into automation. Generating a credit memo is one thing. Making decisions that affect risk exposure is another.

The real challenge is building systems companies can trust operationally. That requires auditability, permissions, human review, governance, monitoring, and clear ownership when systems fail or behave unexpectedly. Once AI starts operating inside regulated workflows, those requirements become unavoidable.

This shift is changing what companies need from technical teams.

The demand is no longer centered on engineers who can simply integrate a model API. Companies need people who understand workflows deeply enough to translate them into reliable systems.

That includes engineers who can work across AI infrastructure, business logic, compliance requirements, and operational systems. It also includes people who understand the industries themselves, because competitive advantage increasingly comes from workflow knowledge rather than raw model access.

As AI vendors move further up the stack, the teams that succeed will not necessarily have the best models. They will understand the work better than anyone else.

The companies that adapt best to this shift will stop treating AI as a feature and start evaluating where human workflows still create defensible value.

Some will move closer to services and operational expertise. Others will specialize in industries where trust and governance matter more than speed. Some will build orchestration layers above the models, while others will become implementation partners for the platforms themselves.

But every company will eventually face the same pressure point: if the model provider can package the workflow directly, what part of the process still belongs to you?

The next phase of AI competition will not be defined only by smarter models. It will be shaped by which companies control the workflows, the operational trust, and the customer relationship around the work itself.

Connect With Other Technology Leaders

If you want to connect with other technology leaders having real conversations about AI and how it is changing business, check out GILD Curated Circuit.

More to come…

✔️ AI companies are racing to own enterprise workflows, not just models - WIRED

✔️ Where competitive advantage lives in the agentic era - Fast Company

✔️ Goldman, Blackstone Partner With Anthropic on AI Services Firm - Bloomberg