AI for the Workforce: Why Context Is the New Competitive Advantage

The biggest opportunity for AI may not be replacing people. It may be making them more capable.

Much of the enterprise AI conversation has centered on productivity: writing software faster, automating repetitive tasks, reducing headcount, or accomplishing more with smaller teams.

But that framing misses another opportunity.

What happens when AI gives the people closest to the work access to information, intelligence, and decision support they never had before?

That question sits at the center of a conversation between Paul Wellons, Founder of Ando; Justin Billig, CEO & Co-founder of Tensure; Dan Rye, Partner & CTO at Tensure; and host Logan Lyles on A Platform for Modern Finance.

Their discussion starts with frontline workers, but it quickly exposes a broader lesson for technology leaders:

AI creates leverage when it has the right context. Platforms determine whether organizations can provide that context safely and at scale.

For financial services and other regulated industries, that distinction is becoming increasingly important.

Watch the full episode of A Platform for Modern Finance

Listen to the full episode on Apple Podcasts or Spotify

‍

Why is context so important for enterprise AI?

AI models may be increasingly capable, but capability alone doesn't mean an AI system understands your business.

It needs context.

A manager might know why one employee works better on certain shifts, which team combinations perform well together, how a local event changes demand, or why an employee consistently chooses one location over another.

Likewise, experienced engineers carry enormous amounts of organizational context: why an architectural decision was made, which service owns a particular function, where an API lives, which workaround exists for a legacy system, or what happened the last time someone changed a particular component.

Historically, much of that knowledge has been trapped inside people's heads.

As Dan explains in the episode, it can also live across Slack and Teams conversations, meeting transcripts, email, source code, commit histories, documentation, and dozens of other systems.

AI creates an opportunity to make that knowledge persistent, accessible, and actionable.

But only if organizations can connect it.

Your company's most valuable AI context is trapped inside people's heads.

Dan explains where institutional knowledge actually lives across engineering organizations and how platforms and agents can make that knowledge available instead of forcing employees to hunt for the one person who knows the answer.

‍

What does AI adoption look like beyond the chatbot stage?

One of the most useful frameworks from the conversation comes from Paul's description of AI's progression:

Analyst → Advisor → Agent

At the analyst stage, AI helps explain what happened.

At the advisor stage, it uses historical information and context to recommend what should happen next.

At the agent stage, AI begins executing parts of the decision itself — within defined boundaries.

Dan sees a similar progression in Tensure's work with technology organizations. The terminology is slightly different, but the pattern is closely aligned:

Consumption → Collaboration → Automation

Organizations initially use AI to search, answer questions, and consume information. They then begin working alongside AI as a copilot. Eventually, specific workflows can be handed off to agents.

The important distinction is that greater capability creates a greater need for context, governance, and clearly defined boundaries.

‍

Why does AI autonomy become a risk-management question?

Once an AI system can take action, the question changes.

It is no longer simply:

Can AI do this?

It becomes:

Are we comfortable letting AI do this?

That distinction matters enormously in financial services.

A coding agent may technically be capable of writing, testing, and deploying software. But a bank or insurer also needs to know what data the agent can access, what actions it can take, who is accountable for the result, where human approval is required, and whether the entire process is auditable.

As Justin points out in the conversation, different organizations — and different workflows inside the same organization — will have different levels of risk tolerance.

That means AI governance cannot simply be a binary choice between autonomy and human control.

Organizations need different levels of autonomy for different levels of risk.

The real question isn't whether AI can do the work.

Justin explains why AI adoption ultimately becomes a question of organizational risk tolerance — particularly in highly regulated industries like financial services.

‍

How is AI changing the software engineering bottleneck?

AI-assisted development creates another unexpected consequence: making engineers faster doesn't necessarily make the organization faster.

It moves the bottleneck.

Dan describes what Tensure is seeing firsthand while working with Ando: engineering output can now move so quickly that the constraints shift toward both ends of the development process.

Before development, organizations have to decide what should be built.

After development, someone still has to determine whether what was built is actually right.

Paul sees the same issue from the product side. When one engineer can produce what previously required a much larger team, organizations risk losing some of the collective reasoning that historically happened around the work.

Teams can ship the wrong thing much faster.

That makes customer understanding, business context, requirements, validation, and tight feedback loops more important — not less.

‍

What can frontline AI teach technology leaders?

At first glance, workforce scheduling and software engineering might seem like very different AI problems.

But they share an important characteristic: valuable context is fragmented.

For a frontline worker, that might include:

• Availability
• Preferred shifts
• Work history
• Location
• Skills
• Career goals
• Family commitments
• Past scheduling decisions

For an engineer, it might include:

• Code
• Documentation
• Architecture decisions
• Tickets
• Slack conversations
• API information
• Deployment history
• Previous incidents

In both environments, the system becomes more valuable when it can learn from previous decisions and make that accumulated context available during the next decision.

Paul calls this a constant learning loop.

Every interaction generates more context. Every accepted or rejected recommendation tells the system something. Every outcome can make the next recommendation more useful.

This is where AI starts moving beyond generic intelligence toward intelligence that understands how a particular organization actually operates.

Every decision creates more context.

Paul explains Ando's constant learning loop and how AI can surface patterns managers couldn't reasonably identify on their own, followed by Justin connecting that concept to workers in financial services.

‍

Why does platform engineering become more important as AI gets more powerful?

Giving an AI agent more autonomy doesn't eliminate the need for a platform.

It increases it.

Without a standardized foundation, an autonomous agent inherits the same fragmentation humans already struggle with: inconsistent tools, unclear ownership, disconnected workflows, permissions problems, missing context, and one-off processes.

Except an agent can operate much faster.

As Justin explains, that makes capabilities such as permissions, observability, APIs, governance, and standardized workflows increasingly important as organizations give agents more autonomy.

The platform becomes the environment in which that autonomy can happen safely.

For financial services technology leaders, this is particularly important. Speed cannot come at the expense of auditability, security, privacy, reliability, or compliance.

The goal isn't simply to put more restrictions around AI.

It's to create a platform where the safe way to operate is also the easiest way to operate.

‍

How do financial institutions add AI guardrails without recreating bureaucracy?

This creates an obvious challenge.

Financial institutions already struggle with approval-heavy processes, tickets, manual handoffs, and governance mechanisms that slow engineering teams down.

Adding AI cannot mean rebuilding the same bureaucracy around every agent action.

Dan recommends starting with risk assessment.

Instead of applying one universal set of controls to every workflow, organizations can classify actions based on risk and apply appropriate guardrails.

A low-risk workflow may require very little intervention.

A workflow involving sensitive data or a consequential production action may require additional controls and human approval.

The important part is making those decisions systematic and automatable.

That creates a model where governance becomes part of the platform rather than another queue engineers — or agents — have to wait in.

AI guardrails shouldn't become TicketOps 2.0.

Dan explains why organizations need different guardrails for different risk levels and how risk assessment can preserve control without rebuilding manual bureaucracy.

‍

Why adoption matters as much as AI capability

The most technically sophisticated platform in the world creates little value if nobody uses it.

The same is true for AI.

Dan points out that adoption is one of the biggest risks organizations face with both platform engineering and AI-enabled products.

If the new workflow creates more friction than the old one, people will work around it.

The answer is not simply more mandates.

Make the new path the easier path.

That principle is central to effective platform engineering: standardized workflows and guardrails work best when developers choose them because they reduce friction rather than because they are forced to comply.

The same lesson applies to AI adoption across the workforce.

‍

Where should leaders start with enterprise AI?

The temptation is to begin by asking:

Where can we inject AI?

Justin recommends starting somewhere else:

What work are we trying to accomplish?

Look at the work itself. Look at the people doing it. Look at the existing process, the desired outcome, and the knowledge required to make good decisions.

Then ask where intelligence can fundamentally change that work.

That distinction matters because AI shouldn't simply be bolted onto processes designed for a different technological era.

Dan makes the same point: organizations that keep the existing process intact and simply insert an AI tool are likely to miss much of the opportunity.

AI adoption may require changing team structures, workflows, responsibilities, feedback loops, and how decisions get made.

Don't ask where you can inject AI.

Dan explains why AI can't simply be bolted onto existing processes, and Justin offers a practical starting point for executives: begin with the work and the outcome, not the tool.

‍

What should financial services leaders take away?

Three principles emerge from the conversation.

First, context is becoming a competitive advantage. Generic AI capability is increasingly accessible. The differentiator is whether an organization can safely connect AI to the knowledge, data, history, and workflows that make its business unique.

Second, autonomy requires architecture. As organizations move from AI assistants toward agents, platforms, permissions, observability, governance, and risk-based guardrails become more important.

Third, adoption is the real test. Whether the user is a software engineer, a bank employee, or a frontline worker, technology only creates value when it makes the work meaningfully better.

For financial institutions trying to modernize without losing control, the goal isn't simply to deploy more AI.

It's to build the organizational context and platform foundation that allow AI — and the people using it — to make better decisions faster.

‍

Frequently Asked Questions

Why does AI need organizational context?

General-purpose AI understands broad patterns, but it does not inherently understand an organization's internal decisions, workflows, systems, customer requirements, risk policies, or institutional knowledge. Giving AI access to relevant organizational context allows it to produce recommendations and actions that are more useful to the specific business.

What is agentic AI?

Agentic AI refers to AI systems that can go beyond answering questions or generating content and take actions toward a defined goal. In the maturity model discussed in the episode, organizations progress from using AI as an analyst, to an advisor or collaborator, and eventually to an agent capable of executing portions of a workflow.

Why is platform engineering important for AI adoption?

Platform engineering can provide AI agents and developers with standardized workflows, permissions, observability, APIs, security controls, and governance. As agents gain more autonomy, these shared foundations help organizations move faster without allowing every team or agent to invent its own operating model.

How should financial institutions govern AI agents?

Rather than applying identical controls to every AI workflow, financial institutions can assess the risk associated with different actions and apply proportionate guardrails. Higher-risk workflows may require stronger controls or human approval, while lower-risk workflows can support greater automation.

Where should executives start with AI adoption?

Start with the work and desired business outcome rather than asking where AI can be inserted. Identify the work being performed, the context required, the existing bottlenecks, and the desired result. Then determine how AI could change the process rather than simply automating one step within the existing process.

‍

Watch the full episode of A Platform for Modern Finance to hear Paul Wellons, Justin Billig, Dan Rye, and Logan Lyles discuss AI for the workforce, agentic AI, platform engineering, and what technology leaders need to do to prepare for increasingly autonomous systems.