# Stonly launches AI agents that run the support playbook humans already use > Source: > Published: 2026-09-17 18:38:13+00:00 # Stonly launches AI agents that run the support playbook humans already use **Dashlane co-founder Alexis Fogel is betting one set of structured guides can govern reps, self-service and autonomous support.** By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff) ยท Published Primary source: [PR Newswire](https://www.prnewswire.com/news-releases/stonly-launches-business-process-agents-to-automate-complex-customer-service-processes-302882341.html) ## Why it matters Customer service AI is moving from generating answers to executing business procedures. Stonly's shared-guide approach could cut maintenance work if it proves reliable on complex, high-risk cases. Alexis Fogel, the Dashlane co-founder who built [Stonly](https://stonly.com/?ref=runtimewire) around a support problem he knew firsthand, launched [Business Process Agents](https://stonly.com/product/business-process-agents?ref=runtimewire) in a [September 17th announcement](https://www.prnewswire.com/news-releases/stonly-launches-business-process-agents-to-automate-complex-customer-service-processes-302882341.html?ref=runtimewire). The AI agents are designed to execute multi-step customer service procedures using the same guides that Stonly customers already give human support representatives. Fogel spent nearly a decade as Dashlane's co-founder and chief product officer. In [his account of starting Stonly](https://stonly.com/blog/starting-stonly-raising-from-accel-founder-thoughts/?ref=runtimewire), he traced the decision to recurring problems with customer instructions at the password manager. Installation and troubleshooting varied by browser, device and operating system, while static help articles struggled to account for each customer's situation. He eventually left Dashlane, he wrote, "to build a new company." Business Process Agents move Stonly from delivering those instructions to executing them. Fogel is betting that the structured support manual can become the control layer for humans, customer self-service and autonomous agents, avoiding a second set of prompts and workflows built solely for AI. ### One guide, three kinds of support Stonly's Guides contain instructions, branches, required checks, customer information, policies and actions. A Business Process Agent can ask a customer for missing details, determine which branch applies, call a connected system and continue until the issue is resolved or human judgment is required. The intended use cases include account changes, troubleshooting, eligibility decisions, billing disputes, intake procedures and policy-driven requests. Stonly says an agent can execute the entire process, complete preliminary work before a human takes over, or pause at a sensitive decision for approval. The same Guide can also remain available to a support representative or appear as customer-facing self-service. If a policy changes, a knowledge manager edits the Guide rather than updating a human procedure and separately rewriting an AI agent's prompts, retrieval rules and workflow configuration. That shared process layer is the product's central argument. Customer service organizations already spend substantial time documenting exceptions, approval rules and escalation paths. Fogel wants Stonly to turn that existing work into executable software without handing every change to an AI specialist or engineering team. The launch carries an unresolved distribution detail. The release's Availability section still contains both options: "available today" and "entering beta with select customers today." Stonly's product page directs prospective buyers to request a demonstration. The materials therefore do not establish whether Business Process Agents are generally available or limited to selected customers. ### Dashlane's support problem became Stonly's product thesis Fogel did not build Stonly alone. Krzysztof Dabrowski, who had worked on Dashlane's product team, joined as a co-founder and helped recruit Stonly's first engineers in Krakow. David Rostan, another former Dashlane colleague who later led sales and marketing at Calendly, joined as a late founder to run Stonly's commercial operation. Fogel described those decisions in [a post about Stonly's distributed founding team](https://stonly.com/blog/building-a-startup-in-the-right-locations-with-the-right-people-and-investors/?ref=runtimewire). Stonly established operations in France, Poland and the United States around the people Fogel wanted to recruit. Dabrowski remained in Poland, while Rostan helped build the go-to-market operation from New York for a product aimed heavily at US support organizations. The financing reflects that transatlantic structure. Stonly disclosed a $3.5M seed round from Accel. [Northzone led a $22M Series A](https://northzone.com/insights/redefining-customer-relationships-through-self-service-our-investment-in-stonly?ref=runtimewire) on January 27th, 2022, with Accel participating alongside Andrey Khusid, Nicolas Dessaigne, Roman Schumacher and former Dashlane CEO Emmanuel Schalit. Stonly has disclosed $25.5M across those rounds, without publishing a valuation. Business Process Agents extend the original interactive-guide thesis rather than replacing it. Stonly began by helping customers and representatives navigate instructions step by step. Fogel is now treating those guides as machine-readable operating procedures that can tell an agent what information to collect, which rule to apply and when to stop. ### Customer service AI is converging on procedures Stonly is entering a market where question answering has become table stakes. Vendors are competing to automate the work that follows an answer: verifying identity, checking eligibility, issuing a refund, changing an account or escalating a sensitive case with the relevant context attached. Intercom's [Fin Procedures](https://www.intercom.com/help/en/articles/12495167-fin-procedures-explained?ref=runtimewire), for example, combine natural-language instructions with branching logic, code, data connectors, simulations and human handoffs. Zendesk [completed its acquisition of Forethought](https://www.zendesk.com/newsroom/press-releases/zendesk-completes-acquisition-of-forethought/?ref=runtimewire) on March 26th, 2026, adding an AI-agent specialist to one of the largest customer service platforms. Sierra [raised $350M at a $10B valuation](https://sierra.ai/blog/theres-an-agent-for-that-and-it-runs-on-sierra?ref=runtimewire) in September 2025 to expand its enterprise agent platform. Stonly's positioning rests on who owns the procedure. Fogel wants support operations and knowledge managers to maintain the instructions in a format they already understand, while the AI executes that same artifact. That approach may reduce duplicated maintenance and make process changes easier to audit. Stonly has yet to publish evidence showing how the approach performs on the complex cases highlighted in the launch. The announcement names no Business Process Agent customers and provides no product-specific resolution rate, error rate, deployment cost or comparison with competing procedure systems. [Stonly's homepage](https://stonly.com/?ref=runtimewire) currently claims 92% response accuracy and a 71% self-service success rate for its wider platform. Those marketing figures are not tied specifically to Business Process Agents, where the harder measurement is whether an agent takes the correct action after navigating several branches and business rules. That proof will determine whether Fogel's support manual becomes an operational control plane or another layer that customers must test and supervise. The underlying bet is grounded in a real support problem: people and AI should not be reading from different instructions when they are expected to produce the same result.