FAQs
Outcomes, guarantees, and expectations
Can you guarantee results?
We guarantee clear deliverables: a production-ready specification, guardrails, metrics, and acceptance tests. Business outcomes depend on your data quality, process ownership, adoption, and how the agent is operated after go-live.
What exactly do you guarantee?
We guarantee delivery of the agreed artefacts and outputs for the stage you purchase (Discovery Sprint, Implementation Sprint, Managed Ops), including acceptance criteria and documentation in a usable form.
What do you mean by “agent”?
An agent is a workflow-bound system that takes an input, applies rules and policy, produces controlled outputs, and escalates when risk is detected. It is not “free-form AI”. It is designed to behave consistently inside a defined process.
Is this just prompts?
No. Prompts are only one part. The core is: workflow scope, decision rules, input/output fields, approved reply templates, escalation policies, acceptance tests, and operational controls.
Is this a chatbot?
No. It’s an agent design (and, if you purchase implementation, an agent build) that operates inside a defined workflow with controlled outputs and escalation rules.
Will it replace my staff?
No. It reduces repetitive work and makes responses consistent. Humans remain responsible for approvals, exceptions, and high-risk decisions.
What if we don’t have clean data?
That’s normal. The sprint identifies what’s missing and provides a practical data plan. If the workflow is too ambiguous or the data too unreliable to automate safely, we’ll say so.
What if our process is messy or undocumented?
Also normal. Discovery exists for that reason. We turn “how it actually works” into a clear spec and operating rules, including what must not be automated.
Process, stages, and what you get
What is the Discovery Sprint?
A short, paid engagement where we define the workflow, boundaries, guardrails, and acceptance tests. The output is a production-ready specification you can implement internally or with us.
What deliverables do we get in the Discovery Sprint?
Typically:
- Workflow map (inputs → decisions → outputs)
- Scope boundaries and “must never do” rules
- Field schema (what the agent extracts / needs / must refuse)
- Escalation policy and handoff format
- Approved response templates and tone rules
- Metrics and acceptance tests (test pack)
- Implementation plan (options, risks, timeline)
Can our internal team build it after Discovery Sprint?
Yes. That’s explicitly supported. The Discovery Sprint gives you the blueprint. Your team can implement it internally, or we can do it in an Implementation Sprint.
What is the Implementation Sprint?
We build the agent according to the specification, wire it into your chosen environment, add logging/auditability, and support UAT (User Acceptance Testing) against the acceptance tests. However, you can choose to build it with your own team or any other software partner — you’re under no obligation to hire us for the implementation.
What is Managed Ops?
Ongoing support to keep the agent reliable: monitoring, prompt/rule updates, change control, incident handling, and performance review. However, you can continue with Managed Ops using your own team or any other provider — you’re under no obligation to engage us for it.
Can we start with Discovery only and decide later?
Yes. That is often the best approach.
What if we want you to implement only part of it?
That’s fine. We can implement a v1 scope and leave the rest as a roadmap.
Systems, access, and integrations
Do we need to give you access to systems?
Not necessarily for the Discovery Sprint. Most work can be done from documentation, exports, and examples. System access becomes relevant for implementation.
Do you build API integrations?
We can, but we only recommend integrations when they add real operational value. Many businesses can run a safe, effective v1 with lightweight tooling and human-in-the-loop controls.
Can this work with WhatsApp, social media, and email?
Yes, but the exact approach depends on whether you want:
- Drafting support (agent proposes replies; humans send them), or
- Automated sending (requires integrations and stricter controls)
We can design both, but most clients start with drafting + routing to reduce risk.
Do you support Microsoft 365 / Google Workspace?
Yes. Discovery is platform-agnostic. Implementation depends on your stack and constraints.
Can we deploy on our own infrastructure?
Yes. We can design for client-managed environments where appropriate.
Can you run it fully hosted by Sienda?
In some cases, yes, subject to scope and security requirements.
Safety, guardrails, and compliance
How do you prevent the agent from “making things up”?
We design “no guessing” rules, structured outputs, and escalation triggers. We also enforce acceptance tests that fail if the agent invents commitments, dates, prices, or operational facts.
Can the agent confirm pricing, bookings, or delivery dates?
Only if your process includes a verified source of truth and you explicitly permit it. Otherwise, the agent drafts a safe message and escalates for human confirmation.
What about sensitive data?
By default, the agent must refuse sensitive personal data and request a safe alternative. We define and enforce “must never request/store” rules.
Can the agent handle complaints or refund threats?
Yes, but usually via escalation: safe holding reply + structured handoff to a human.
Do you provide legal/compliance advice?
No. We design operational guardrails and safe behaviours, but legal and compliance decisions remain yours. We recommend involving your legal counsel where needed.
Is it GDPR compliant?
We operate in compliance with UK GDPR and, where applicable, EU GDPR. We minimise data, control retention, and avoid using client data for model training.
Confidentiality and data handling
Are our details confidential?
Yes. We operate under strict confidentiality and treat all client information as confidential business data.
Do you use client data to train AI models?
No.
Can you publish our case study?
Only with explicit written permission and within agreed disclosure limits.
Do you offer white-label delivery?
Most of our clients ask for white-label delivery and strict confidentiality — it’s always been one of our defining traits. We routinely anonymise client data and keep everything we learn from their processes and information under the tightest confidence. So, yes, we can deliver without Sienda branding and remain invisible.
Costs, pricing, and commercial terms
How is pricing structured?
Typically by stage: Discovery Sprint fee, Implementation Sprint fee, and a monthly Managed Ops retainer (if chosen). Pricing depends on workflow complexity and risk profile.
Do you offer fixed pricing?
Often yes for Discovery. Implementation and Managed Ops may be fixed or blended depending on scope clarity.
What third-party costs should we expect?
Potentially: hosting, model usage, logging/monitoring tools, and any platform subscriptions. We’ll make those explicit before implementation.
Can we cap model spend?
Yes. We can implement budget guards, request limits, and caching strategies.
What if we stop Managed Ops?
You keep your deliverables. After handover, ongoing changes and reliability become your responsibility unless you retain us.
Timeline and readiness
How long does it take?
Discovery is typically short (days to a couple of weeks, depending on availability and workflow complexity). Implementation depends on integrations and scope.
What do you need from us to start Discovery?
Usually:
- A clear workflow to target (one is enough for v1)
- Examples of real messages/requests (anonymised if needed)
- Current response templates or policies (if any)
- Who owns the process internally (decision-maker + operator)
What if we’re not sure what workflow to automate?
We can help choose the best first workflow based on volume, risk, and measurable impact.
Quality control and testing
How do you test the agent?
We create acceptance tests (a test pack) that the agent must pass. These test for safety, correctness, tone, and escalation behaviour.
Can we add our own tests?
Yes. In fact, we encourage it. Your edge cases become part of the test pack.
What does “production-ready specification” mean?
It means a document your team can implement without guesswork: fields, rules, templates, boundaries, escalation, metrics, and pass/fail tests.
Ownership, IP, and handover
Who owns the deliverables?
Client-specific deliverables are provided to you for internal use under the Terms. Sienda retains background IP and general methods.
Will we be locked in?
No. Discovery outputs can be implemented by your internal team. Managed Ops is optional.
Can you train our staff?
Yes. We can provide training on operating the agent, updating templates, reviewing outputs, and managing drift safely. This is priced separately, depending on the exact scope of what’s required to train your staff.
Common “what if” questions
What if the agent makes a mistake?
That’s why we design guardrails, escalation, and human approval points. Managed Ops further reduces risk through monitoring and controlled updates.
What if our workflow changes?
We update the specification (and implementation, if included). Change control is part of how agents stay reliable.
What if we want multi-language replies?
Possible. We treat multilingual support as a defined requirement with its own acceptance tests.
What if we want the agent to handle multiple workflows?
We usually start with one workflow for v1, then expand. Multiple workflows are best handled as separate “modules” with their own tests.
What if we need 24/7 coverage?
That’s an operational requirement. We can design for it, but it affects tooling, escalation, and support terms.
Getting started
What’s the best first step?
Start with the pre-sales enquiry form. If it looks like a fit, we’ll suggest the best v1 workflow and the appropriate stage (Discovery first, then decide on implementation).
Can we see a demo first?
Yes — we can share a private demo link to a reference triage agent. It’s a demonstration of behaviour (guardrails, escalation, audit trail), not a tailored client build.
Where do we request the demo link or pricing?
Visit the book page here at sienda.co.uk and use the pre-sales enquiry form, or book a consultation via the booking page.
Contact us if you have further questions