AWRA OpsHub Search
AWRA AI Assistants

Two copilots. Neither one makes things up.

One copilot rides along the public site; the other lives inside your workspace. Both answer from AWRA's own knowledge and point you to the exact page you need — and the in-app one also answers questions about your own data, under your permissions, with figures no AI model has touched.

Meet both

Grounded on AWRA's own answers · Your data under your permissions · Deep-links you into the product · Escalates to a human

2 copilotsone for the public site, one inside your workspace
0 made-up factscurated knowledge, or figures read straight from your workspace
Your permissionsdata answers only where your role already sees the figure
Human handoffknows when to pass you to a real person
The idea

Most AI chat is confidently wrong. This one would rather point you to the truth.

A support bot that invents a feature you don't have, or quotes a price that isn't real, does more damage than no bot at all. It erodes trust at exactly the moment someone is deciding whether to rely on you. So the AWRA assistants are built the other way around: instead of generating plausible-sounding text, they find the right answer in AWRA's own knowledge and hand it to you, with the page to prove it.

They understand what you typed — expanding synonyms and fixing common misspellings — then match it to curated answers and indexed help content, score how confident they are, and respond accordingly. High confidence gets a direct answer. Lower confidence points you to a resource. And when it's genuinely a question for a person, they say so and connect you. An assistant that knows the limits of what it knows is worth far more than one that pretends it has none.

Meet the two

Same principles. Two very different jobs.

AwraIQPublic website copilot

The assistant for anyone exploring AWRA. It searches the site's indexed knowledge first, then falls back to curated intents — so a visitor can ask about the product in plain language and get a real, sourced answer.

  • Answers product, security, onboarding, integrations and use-case questions
  • Deep-links to the exact feature and resource pages
  • Offers suggested follow-up questions to keep exploring
  • Captures your details and hands you to the team — with a WhatsApp shortcut
Persona “AwraIQ” · lives on every public page
AWRA AssistantIn-app copilot

The concierge for logged-in teams. Ask how to do something and it explains the first steps, then drops you straight onto the right screen. Ask for a number and it reads it from your workspace. It even knows the page you're on and tailors its suggestions to it.

  • Explains how to complete common tasks across every module
  • Deep-links you into the exact in-app page to act
  • Aware of where you are — boosts answers relevant to your screen
  • Answers everyday questions from your own data — approvals waiting, low stock, overdue invoices, till sales, late orders, your tasks, tickets and alerts — under your own permissions, never reworded by an AI model
  • Never acts for you — it reads and guides, you decide
Persona “AWRA Assistant” · launcher “Ask AWRA” · signed-in users Get access
Ask about your own numbers

The in-app assistant reads your workspace — and keeps the AI away from the figures.

Ask “what’s low on stock?” or “sales this week?” and the AWRA Assistant answers from your own records, not from a help page. Three rules hold for every one of those answers, and they are why you can repeat the number in a meeting.

1

Your permissions first

Each question is tied to the access that shows the same figure on screen. Without it, the assistant says your role does not include it and reads nothing — it is never a way round a role.

2

No AI in the loop

The figure is the query result, worded by a fixed template. It is never passed to a language model, because a model asked to restate a count can round it, reword it or invent one.

3

Currencies are never added together

A money total is given in your largest currency, with the rest disclosed beside it — never converted and summed into a figure nobody can check.

The questions it answers from your data

Ask something likeIt readsYou need
“What needs approval?”Every pending item in the approvals inbox, by moduleDashboard access
“What’s low on stock?”Items at or below their reorder pointAccess to items
“How many safety boots do we have?”The stock of the items matching that nameAccess to items
“Which invoices are overdue?”Overdue customer invoices, the total and the oldestAccess to invoices
“Sales today?” / “this week?” / “this month?”Till sales for the period, and says it excludes invoiced salesAccess to POS
“Any late purchase orders?”Open orders past their delivery date, with supplierAccess to purchase orders
“What are my tasks?”Your own open tasks, overdue and due todayNothing extra — they are yours
“Open tickets?”Tickets assigned to you; the whole queue if you can see all ticketsYour own tickets need nothing
“Any alerts?”Active Smart Alerts with their owner or acknowledgementDashboard access

Asked “how do I…”, the assistant still explains the steps and deep-links you to the screen — a how-to question goes to the guide, not to the data. It reads and never changes anything, and links it shows you are filtered to the screens your role can open.

One grounded brain

How a question becomes a trustworthy answer.

Every message runs the same disciplined path — understanding, grounding, scoring, answering, and escalating when needed.

1

Understand

Your question is normalised, synonym-expanded and de-typoed so “costings” or “patner” still land on the right topic.

2

Ground

It searches AWRA's own indexed knowledge and curated intents to find the best-matching real answer.

3

Score

A confidence score decides whether to answer directly, merely suggest a resource, or hand off.

4

Answer

You get the curated answer plus deep links to the exact page — optionally AI-polished, always grounded. A figure from your data is never polished.

5

Escalate

Unsure or sensitive? It routes you to a human — a lead handoff on the site, WhatsApp in the app.

The one rule the AI layer cannot break

When the optional AI language layer is switched on, it works under a strict instruction: answer using only the provided snippets — do not invent features, pricing or claims. It can make a grounded answer read more naturally, but it cannot manufacture facts. If every model is disabled or unavailable, the assistant simply serves the curated answer as-is. And an answer built from your own data never reaches the AI layer at all. Grounded first, polished second — never the other way around.

Honest by design

It tells you how sure it is — and stays in its lane.

High confidence

A strong match — the assistant answers you directly, with sources.

Suggested resource

A partial match — it points you to the most relevant page to read.

Human follow-up

Not confident enough — it recommends connecting you with a person.

Safe boundaries on the topics that matter

Some questions shouldn't get a breezy bot answer. On these, AwraIQ deliberately gives careful guidance and steers you to the team instead of overpromising:

Pricing Legal & privacy Timelines Integrations Enterprise promises
Under the hood

Small footprint. Serious manners.

01

Knowledge search

AwraIQ indexes the public site — features, docs, academy and blog — and searches it to answer from the real thing.

02

Deep links

Answers come with links straight to the relevant page — a feature page on the site, or a module screen in the app.

03

Suggested questions

Each reply offers smart follow-ups, so people discover what to ask next instead of staring at a blank box.

04

Context-aware in-app

The AWRA Assistant reads the page you're on and boosts answers relevant to it — help that fits where you are.

05

Human handoff

On the site it captures your details into a lead form; in the app it offers a WhatsApp shortcut to support.

06

Rate-limited & logged

Both assistants are throttled and log each turn — intent, confidence and outcome — so the experience keeps improving.

07

Optional AI polish

When enabled, a grounded language layer rephrases answers for clarity — strictly from AWRA's own content.

08

Data-safe

The in-app assistant reads only what your own role can already see, refuses where you lack access, never sends a figure to a language model, and takes no actions.

09

WhatsApp shortcut

Prefer a person? A WhatsApp quick-contact is one tap away from either assistant.

Not to be confused with

The assistants help you ask. Awra Insights analyses your data.

Questions, answered

AWRA AI Assistants FAQ.

Is this a ChatGPT-style chatbot?
No. The assistants are grounded on AWRA's own knowledge and curated answers — not a general-purpose chatbot. They won't invent features, pricing or claims. An optional AI language layer can rephrase answers for clarity, but only from AWRA's own content, never to generate new facts.
What's the difference between AwraIQ and the AWRA Assistant?
AwraIQ is the public copilot on the website, answering visitor questions and capturing leads. The AWRA Assistant is the in-app copilot for logged-in users, explaining how to do tasks and deep-linking into the right module.
Does the in-app assistant read or change my data?
It reads, and never changes. For a fixed set of questions — pending approvals, low stock, a named item’s stock, overdue invoices, till sales, late purchase orders, your open tasks and tickets, active Smart Alerts, and how much of your plan is in use — it reads the figure from your workspace under your own permissions, and says so if you lack access. It never sends those figures to a language model, and it takes no actions for you — you always stay in control.
Can the assistant show me figures my role cannot see?
No. Each data question is tied to the access that shows the same figure on screen. Without it the assistant says your role does not include that access and reads nothing, and the links it offers are limited to screens you can open.
What happens when it can't answer?
It's honest about it. On the website it captures your details and hands you to the team (with a WhatsApp shortcut); in the app it points you to WhatsApp support. It escalates rather than guessing.
Do assistant answers count against my plan’s AI requests?
Only answers written by the AI language layer count against the AI requests a month your plan includes. Figures read from your data and curated answers do not. When the month’s requests are used up, the assistant keeps answering from curated knowledge until the 1st, and an admin can add packs of 500 requests.
How does it handle pricing or legal questions?
Those topics have safe boundaries. Instead of committing to figures or terms, the assistant gives careful guidance and directs you to the team, so nothing is over-promised.
Is it the same as Awra Insights?
No. Awra Insights is a separate analytics and forecasting feature that summarises your own operational data. The AI Assistants are conversational helpers for finding answers and getting around the product.
Help that tells the truth

Don't take our word for it — ask the assistant.

AwraIQ is live on this page right now. Ask it about pricing, security, integrations or a demo, and see a grounded, honest answer for yourself.

Get started free