Stonly launched Business Process Agents (BPAs) on September 17, 2026 - AI agents that automate complex customer service processes by following the same guides and standard operating procedures human support teams already use. The move targets the hardest part of support automation: high-value issues where the right resolution depends on customer context, company policy, and step-by-step decision logic, not just a single answer.
How Business Process Agents work
BPAs run on Stonly Guides, structured workflows that define the steps, rules, required information, and actions behind a support process. Instead of requiring teams to translate their procedures into separate AI prompts, retrieval rules, and workflow configurations, BPAs execute the same guide a human agent would follow. They can ask for information, determine what happens next, take action through connected systems, and continue until the issue resolves or a person needs to step in.
This means a single guide works across three modes: a human agent following it, a customer using it for self-service, or an AI executing it autonomously. When a process changes, the team updates the guide once. The new version is immediately available everywhere it is used - no separate maintenance for human and AI versions.
What BPAs can automate
The agents handle processes that cannot be reduced to a simple question-and-answer pair. That includes troubleshooting, eligibility decisions, account changes, policy-driven workflows, intake procedures, and escalations - the "it depends" cases that consume meaningful agent time. BPAs follow branches, apply decision logic, check customer information, and enforce business rules at each step.
Teams can also choose how much autonomy to give the agent. A BPA can complete an entire process independently, handle part of it and ask a human to approve a key decision, or collect information and procedural work before handing the case to an agent with full context and the next step ready. This avoids forcing every interaction into a binary choice between fully manual or fully autonomous service.
Why the duplication problem matters
Most support organizations already have documented SOPs for their complex processes. When they try to automate those processes with AI, they end up recreating the same logic in a different form - prompts, agent instructions, retrieval rules, workflow triggers. Over time, they maintain two versions of the same process: one for people and one for AI. That duplication creates governance risk, slows down changes, and limits how much work teams are willing to hand over to automation.
"Without Stonly, teams end up recreating the same processes for AI that their support teams already know how to follow," said Alexis Fogel, Founder and CEO of Stonly. "The opportunity is much bigger than making prompts better. It is giving AI the ability to follow the same processes the business already trusts its people to follow."
BPAs remove the duplication by making the guide the single source of truth. AI follows the actual process, not a separately maintained approximation of it. That lets automation extend to processes previously considered too complicated or too risky to hand off.
For support professionals looking to build skills around this kind of automation, structured learning paths like the AI Learning Path for Technical Support Specialists cover the intersection of AI agents and support workflows. The broader AI for Customer Support resources also track how tools like BPAs are changing what support teams can automate.
Why this matters for customer support professionals
BPAs shift the automation ceiling. The processes that consume the most agent time - the multi-step, policy-heavy, context-dependent ones - are the same processes that have been hardest to automate. Stonly's approach means support teams do not need to become AI engineers to get AI working on those cases. They define the process once, in a format they already manage, and AI follows it. For support leaders under pressure to cut costs without sacrificing service quality on complex issues, that changes which problems are realistic to automate.
Your membership also unlocks:
