NIA understands the request, retrieves authorised context, recommends or performs the next approved action, and records what happened. It is an agent platform you configure by organisation, department and role, not a chatbot bolted onto a website.
Understand. Decide. Act with confidence.
You do not have to take our word for any of this. The assistant in the corner of this page is NIA running in consultation mode, and every stage below is what she actually does.
A new project, advice, a partnership discussion or support for a system you already run. The conversation branches from there.
An association exploring a member platform does not get the same questions as a business that wants a website redesign. She asks what is relevant and skips what is not.
The right service or combination, and why. She works from an approved NSDS knowledge base, so she cannot invent a price, a capability or a result.
What an assessment, a consultation or a proposal involves, what is included and what is not, and indicative durations clearly labelled as indicative.
She assembles a summary, shows it to you in full, and lets you correct it. Nothing reaches the team until you press send.
Your summary, the answers behind it and the open questions, so nobody asks you to repeat the conversation. Transcripts are staff-only and expire after 180 days.
| NIA is | NIA is not |
|---|---|
| A modular agent platform configurable by organisation, department and role | A generic public chatbot with unrestricted access |
| A governed interface to knowledge, workflows and approved software tools | A replacement for management judgment, regulated advice or accountable staff |
| A product family NSDS can deploy, manage and improve continuously | A one-time website widget disconnected from business operations |
| A measurable operational capability with permissions, logs and service levels | An opaque automation that acts without traceability |
Eight stages, each with a control. The controls are the product, not an afterthought bolted on once the model is working.
Accepts text, voice, file or system event.
Channel and identity validationIdentifies intent, task, urgency and missing information.
Prompt safety and ambiguity checksRetrieves authorised knowledge and records.
Source, tenant and role restrictionsSelects an answer, a recommendation, a workflow or a tool.
Policy engine and confidence thresholdRequests approval for consequential actions.
Human in the loopExecutes through an approved connector or workflow.
Least privilege, transaction controlsLogs inputs, sources, decisions, approvals and results.
Audit and retention policyFeedback and outcomes inform reviewed updates.
Evaluation and controlled releaseFive discrete modes, each with its own risk rating. This is a configuration choice per use case, not a ladder NIA climbs on its own.
Answers and explains only.
LowProposes a next action. Cannot execute.
Low to moderateDrafts or stages an action for review.
ModerateExecutes after an authorised person confirms.
Moderate to highExecutes pre-approved low-risk actions autonomously.
Controlled, repeatablePeople talk to NIA. Behind the conversation, role-based specialists carry the work, each scoped to the knowledge and tools that role is permitted to use.
Knowledge and navigation. Finds the policy, the document, the record or the right person.
Customer service. Status lookups, eligibility, intake, ticket creation and hand-off.
Lead qualification, discovery intake, proposal preparation and follow-up discipline.
Procedure guidance, work orders, checklists, scheduling and incident triage.
Invoice questions, document collection, variance explanation and deadline reminders.
Obligation lookup, evidence collection, control checks and remediation tracking.
Briefings, meeting preparation, decision logs and the narrative behind the numbers.
Every use case carries a named human checkpoint. The checkpoint is defined before the agent is built.
| Domain | What NIA does | Human checkpoint |
|---|---|---|
| Customer service | FAQs, eligibility, quotation intake, order and status lookup, ticket creation | Exceptions, refunds and commitments |
| Sales | Lead qualification, discovery intake, proposal preparation, follow-up reminders | Pricing, negotiation and the final offer |
| Operations | Procedure guidance, work orders, checklists, scheduling, incident triage | Safety-critical or unusual work |
| Finance | Invoice questions, document collection, variance explanation, deadline reminders | Posting, filing, payment and advice |
| HR | Policy navigation, onboarding, leave guidance, training support | Employee relations and decisions |
| Compliance | Obligation lookup, evidence collection, control checks, remediation tracking | Legal interpretation and sign-off |
| Executive support | Briefings, meeting preparation, decision logs, KPI narratives | Strategic decisions and external commitments |
These five boundaries are absolute. They are not configurable, they are not waived for a pilot, and no feature we build will contradict them.
Single sign-on, multi-factor authentication, session controls and verified service accounts.
Role and attribute based, least privilege, with read, prepare, approve and execute rights held separately.
Data, indexes, configuration, keys and logs separated per organisation.
Encryption, minimisation, purpose limitation, retention, deletion and transfer assessment.
Injection defence, connector allowlists and explicit transaction boundaries.
Tamper-resistant logs of inputs, sources, decisions, approvals and results.
Four tiers. We charge for business capability and managed outcomes rather than for messages alone, and usage is shown transparently. Final pricing is set per engagement after discovery.
Small organisations. One agent, curated knowledge, web and portal channels, basic analytics.
Multiple agents, standard integrations, approval workflows, admin console, enhanced analytics.
Single sign-on, advanced governance, custom integrations, dedicated environments, managed service.
APIs and in-product experiences, shared components and usage controls for software you already run.
No price is published for NIA. Figures are set per engagement after discovery and shown in full before you commit.
NIA is the intelligence layer across the rest of what we build: consulting engagements, managed technology, websites, software platforms, hosting, Privacy Hub, JATLAS and ClearBook. It can be bought on its own, embedded in an NSDS application, or built into a transformation programme. The contracting entity remains Next Step Group Limited, trading as Next Step Digital Solutions.
She is live on this website as our consultation and intake assistant, which is the first production deployment. The wider platform is in controlled pilot: we are running it inside NSDS as a knowledge and operations assistant so the grounding, approvals and audit trail are proven on our own work before we deploy it for a client. Pilot engagements are open to selected organisations now.
Yes, and she says so on first contact. She runs from an approved NSDS knowledge base rather than answering freely, which is deliberate: it is what stops her inventing a price or a capability. When a question falls outside that knowledge base she says so and passes it to a colleague with your question attached, rather than guessing.
There is no published price, and we will not invent one. The pricing architecture has five components: a discovery and readiness fee, an implementation fee covering knowledge preparation, integrations, workflows and governance, a monthly platform fee by tier, a usage fee, and a managed service retainer. The figures are set per engagement after discovery, and you see them before you commit.
Because it answers from your approved sources rather than from open-ended generation, it shows which source it used, and it is explicitly prohibited from presenting uncertain information as fact. Where it does not know or does not have permission, it says so and tells you who can help. That behaviour is tested before release, not assumed.
No. Each use case is configured to one of five modes, from answering only through to executing pre-approved low-risk actions. Anything financial, legal, employment-related, safety-related, regulated or irreversible requires a named person to confirm, and the approval is logged with the action.
No. Tenant isolation covers data, search indexes, configuration, keys and logs. Cross-tenant exposure is one of the five boundaries the platform is built to prevent rather than a setting an administrator can get wrong.
No, and that is rather the point. NIA connects to what you already run: Microsoft 365 or Google Workspace, your CRM, help desk, accounting, inventory, HR and payment systems, through an allowlist of approved connectors with least-privilege access.
NSDS is, for the platform, the grounding and the controls, under the service levels in the agreement. Your organisation remains accountable for the decisions your people make. NIA is built to make that line clear rather than to blur it, which is why every consequential action names an approver.
Bring the process that consumes the most time answering the same questions. We will tell you whether it is a sensible first agent, what it would take to ground it properly, and what the checkpoint should be.
Powered by Next Step Digital Solutions
