How is an AI agent different from an RPA script?+
An RPA script records a fixed click path through a user interface; it breaks when the interface changes and it has no way to handle a case it has not seen before. An AI agent reasons over the inputs, calls tools against your systems of record, follows written guardrails about what it can and cannot do, and escalates to a named human when a case falls outside its guardrails. It does not need to be rebuilt every time your ERP ships a new release.
How is this different from Zapier, Make, or Workato?+
SaaS automation platforms are great at wiring public APIs together, but they cannot write to the specific fields in your ERP, warehouse, or billing system where real operations work happens, and the audit trail they produce is split across their platform and the systems they call. A Deploris AI automation runs against your systems of record with named writes, a 48-hour reversibility window, a single per-case audit trail, and EU-region hosting if you need it.
How long does it take to go live?+
A first working version lands on your staging environment inside 4 to 6 weeks. A full production rollout for one process, including a two-week shadow run, typically completes in 3 to 6 months. Additional processes join the stack on the same two-week cadence once the operating model is in place.
Where will the agent and its data live?+
By default in the Frankfurt (EU) region of a hyperscale cloud under a signed data processing agreement, with the option to run in your own cloud account. We do not route data through US-only services anywhere in the pipeline when EU-region hosting is in scope. The model provider, embedding provider, and vector store are named up front in the written scope.
What happens when the agent is wrong?+
Every write the agent makes is reversible for 48 hours, every tool call is logged, every decision is explainable from the audit trail. If an exception is misclassified, the human reviewer corrects it, the correction feeds back into the calibration data, and the guardrails are updated in the next two-week increment. There is no case where an agent decision is unrecoverable and no case where leadership cannot see what the agent did.
Where do humans stay in the loop?+
The written scope names every approval gate before any code is written: amounts above a monetary threshold, writes to specific sensitive fields, cases flagged with low confidence, counterparties on a watch list. Humans stay on the risky moments, not on every moment. The target is not zero humans; the target is zero humans on work that software should have done.
What is the service-level agreement after launch?+
15-minute median first-response on priority-one incidents, 99.9 percent monthly uptime target on the systems we operate, 48-hour reversibility on any shipped change with full audit trail, named on-call engineer, weekly patching window. Specific numbers are written into every engagement contract, not a marketing page.
How do you handle GDPR, DPA, and EU AI Act requirements?+
A data processing agreement is signed before build begins and covers every system in the pipeline. EU-region hosting is the default; sub-processors (model provider, embedding provider, vector store) are listed and reviewed quarterly. Every AI automation is inventoried against the EU AI Act risk tiers and ships with the documentation that act requires: purpose, inputs, outputs, human oversight, monitoring, and incident procedure.
Which processes are a good fit for this?+
Repeatable multi-step work with a defined start and end, written-down rules for most cases, named systems of record, and a measurable exception rate: order triage, invoice reconciliation, back-office routing, support ticket pre-triage, data entry from inbound documents, master data cleanup, vendor onboarding checks, SKU and catalog hygiene. If the process lives in one person's head and has never been written down, we start by writing it down.
Who owns the code, the prompts, and the data?+
You do, in all three cases. The source repository, the prompt library, the infrastructure-as-code, and the data are yours from day one. Deploris operates the system under contract; nothing in the stack is a Deploris black box, and you can take operations in-house or move to a different operator at any time without a rewrite.
Can you work with our in-house engineering team instead of replacing them?+
Yes. On many engagements Deploris leads the build and your in-house team learns the stack alongside the engineers, taking over selected agents over time. The written scope names which agents Deploris owns, which the client team owns, and which transition during the engagement.
Who is Deploris and why should we trust you to run our operations?+
Deploris is a US-registered engineering firm (Colorado LLC 20268259375) with a DACH delivery team, built as the mid-market answer to the gap between dev shops that write software and MSPs that run infrastructure. Of 40 mid-market firms benchmarked across the US and DACH, we are the only team that covers both halves under one contract. The person on your on-call escalation is the same person who wrote the agent that paged.
