Presented by Appian

Artificial intelligence is creating a new reason for government agencies to confront an old problem: applications that cost too much to maintain and no longer support the way people need to work. Matt Calkins, Chief Executive Officer & Co-Founder of Appian, says the opportunity extends beyond updating software. Agencies can use modernization to reshape the systems and processes that will support their next stage of AI adoption.

Speaking with Francis Rose on Beyond the AI Pilot: Turning Investment into Mission Outcomes, presented by Appian, Calkins describes a combination of pressures making legacy modernization more urgent. Aging applications can be difficult to use, require extensive training, and trap information in separate systems. He also argues that evolving AI capabilities are increasing the security risks associated with leaving outdated applications in place.

The starting point, he says, is focus. Agencies need an inventory of applications that require attention and a plan for addressing them. A long list becomes useful when leaders establish priorities and decide what the future environment should accomplish.

Calkins cautions against treating that plan as a series of one-for-one replacements. An agency with many separate applications may be able to consolidate functions, simplify access, and create a more consistent experience for employees. Shared information and common training can become part of the modernization outcome.

That enterprise perspective becomes particularly relevant as AI agents begin interacting with systems in less predictable ways. Calkins describes agents that search across applications, consider information, and act on what they find. Their behavior creates new demands for an environment originally designed around more repetitive exchanges between systems.

Agencies therefore need to consider how agents identify authoritative information, obtain current data, and observe security restrictions. A connection to information is only useful if the system can access it appropriately and reliably. The architecture surrounding AI becomes central to whether people can trust it with consequential work.

Calkins outlines four capabilities Appian emphasizes in addressing that challenge. The first is a deterministic layer that brings predictable process structure to probabilistic technology. The second is secure, efficient access to data, which he connects to Appian’s data fabric. The third is governance, including tracking, guardrails, escalation, and validation.

The fourth is matching work to the agent or model best suited to perform it efficiently. Calkins argues that token consumption should be managed carefully. Sending every task to the most advanced model can increase expense without being necessary for the work at hand.

Together, those capabilities frame AI adoption as an operational design challenge. Agencies must decide how work flows, what information supports it, where controls belong, and how resources are allocated.

For leaders trying to move beyond pilots, Calkins’s message is practical: modernization and AI strategy should reinforce each other. Better applications can make data more useful, shared processes can reduce fragmentation, and clear controls can help agencies extend AI into work that carries greater responsibility. The intended result is an enterprise prepared to turn technical capability into sustained mission performance. That requires attention to both the underlying systems and the everyday decisions that determine how effectively people and agents work together.

3 Key Takeaways

• Modernization should consolidate and improve applications rather than reproduce existing silos.
• Mission-critical AI needs predictable processes, secure data access, governance, and appropriate task routing.
• Assigning work to the right agent or model can help agencies manage AI costs as usage grows.