An enterprise automation platform for agents and workflows

Terms such as business process automation, intelligent automation and hyperautomation overlap and are used by analysts and vendors for slightly different things. Strip the labels and the buyer’s question is the same: how do we automate work across many teams without losing control of it?

A team automating its own tasks needs a builder and some connectors. An enterprise needs those plus answers to questions that only appear at scale. Who is allowed to build? Who approved this automation to touch the finance system? Can we prove what ran last March? What happens when the person who built it leaves?

AI agents sharpen these questions. An agent can decide, so the automation has a kind of discretion that a fixed workflow never had. The platform must be able to bound that discretion and show it.

Integrations: Studio lists 100+ tool integrations and an API for custom systems. That is a smaller catalogue than the largest integration platforms, so check it against your list before anything else.

Identity: role and tenant controls are available. SAML single sign-on and SCIM provisioning are in development and not generally available. If they are a gate for you, say so early.

Agent authority and approvals: Studio has human-approval steps in workflows, and Nexus applies policy to agents while they run. Each agent has an identity, permissions and limits, and consequential actions can require a named approver. The platform describes five levels of autonomy, so an agent can start by recommending and be given more room as its record supports it.

Audit: each run records traces, logs and token-cost telemetry per step, and configurable data-retention policies are designed for, not shipped. The platform’s model is to trace the chain from data to model to agent to decision to action to outcome.

Change control: workflow and chatflow versions can be rolled back and chatflow versions A/B tested in Studio. Golden sets, regression runs, release channels and gradual rollouts are in development.

Deployment: Studio is a managed service, with customer data at rest in AWS Ireland. Dedicated or private deployment is designed for and scoped with Swfte as an engagement. It is not self-serve, and there is no self-serve self-hosting.

Cost: Free (10 workflows, 5 agents, 2,000 executions per month); Pro $39; Team $99 (3 users included); Enterprise custom. Studio meters executions. Model usage goes through the Connect gateway, which provides budgets and alerts.

Assurance: Swfte does not hold a SOC 2 report or an ISO 27001 certificate and does not sign HIPAA Business Associate Agreements today; a SOC 2 Type I audit is in preparation, and the trust page has the current status. We do not claim that using Studio makes your processes compliant; that depends on your use case, jurisdiction, deployment and configuration.

These are patterns, not statistics. We have not measured their frequency.

• Citizen-built automations with no owner. Someone builds something useful, leaves, and nobody knows what it touches. Require an owner and a purpose on every automation.
• Credentials shared across automations. When one leaks, everything it touched is exposed. Prefer an identity per agent and per workflow.
• Approvals bolted on afterwards. A process designed without a human checkpoint rarely gains a good one later. Design them in.
• Licences priced per run when the process loops. Retries, branches and agent steps multiply the count. Ask for a worked example at your volume.
• No exit plan. Ask on day one how definitions and history are exported.

Start with a process that has a clear owner, a measurable result and low consequence if it goes wrong, such as drafting replies or routing requests with a person approving each one. Run it at the assist level, where the agent recommends and a person decides, and keep the record.

Move to approval-by-exception only after the record shows the agent has been right on real volume. Then add the next process. At each step, the question is what the agent is allowed to do and who said so, not what it is capable of.