AI-BI Issue Desk

A vision for Nestlé Vietnam AI Reinvention, from AIT-SGN

Demo on synthesized data. AIT-SGN is not affiliated with Nestlé.

The vision

Every morning, every team at Nestlé Vietnam knows what changed in its numbers, why it changed and what to do about it, in the group chat the team already works in.

No platform of its own. You @tag the AI and message it directly in your team's group chat, alongside your teammates. Everyone sees the question and the answer.

Decisions on today's knowledge, not on last week's report.

Each team gets the signal on the day it happens, the reason behind it and a suggested next step, without opening a dashboard or waiting for an analyst.

Every answer comes from three things that are Nestlé's own:

Live data
the company's own numbers, as they stand today
Business definitions
one meaning for every number, in every function
Working context
who acts, and how the team handles each kind of issue

The problem

Fresh data, stale knowledge

The data is rarely the problem: in most companies it is updated every day. The problem is what happens between the data and the people who decide.

1 · Fresh data, stale knowledge

Data is updated every day. The signal in it, what moved and why, arrives much later, after someone has consolidated and analysed it. Decision-makers work on knowledge that is already old.

2 · Dashboards are built, then neglected

A dashboard is hard to read for anyone who did not build it, so it is opened less and less. Going deeper is harder still: asking “why” means a request to an analyst.

3 · Each function has its own view

Each function has its own dashboards and data views, which other functions find hard to read. The links between them do not show, so functions cannot easily work on the same issue together.

What is missing: the signal on the day it happens, a way to go deeper in plain words, and one meaning for every number, so that functions can work on the same issue together.

The idea

One company. Many slices. A snapshot of each, every morning.

AI-BI sees the company through different business slices: by function, category, brand, and more. Each slice is understood through its own business ontology: the concepts, metrics and relationships that define how that part of the business works. Every morning, AI-BI uses that ontology to capture a snapshot of the slice.

By function

Sales

Slice ontology

revenuepromotionchannelpipelinetarget

Snapshots

By category

Coffee

Slice ontology

category growthshelfpricecompetitorsconsumer

Snapshots

By brand

Milo

Slice ontology

availabilitydistributionsharemediamargin

Snapshots

Governed business ontologyshared definitions, linked concepts, approved meanings

  1. 1Slice the company
  2. 2Interpret each slice through its ontology
  3. 3Snapshot each slice every morning

Each morning's snapshot is compared with the slice's own past: yesterday, last week, last month, last quarter, last year. That is where an issue shows, its impact is estimated, and a deeper question or a suggested action follows.

In the group chat

One channel, one morning: the Milo brand

The images are illustrations, not screenshots of a live workspace. The figures are computed on synthesized data; distributor names are invented.

1. Snapshot, compare, raise, size, then a deeper question. The AI speaks first in the team's group chat. A teammate @tags it, and the whole channel reads the answer.

Illustration of a Slack channel: AI-BI posts the Milo morning snapshot, a teammate @tags it to ask why, and AI-BI answers in the thread.

2. One card per issue. Turn a complex situation into a single, clear card: what happened, how big it is, why it happened, and what to do next.

Illustration of an issue card for the Ho Chi Minh City sell-out gap, with its size, its cause and a suggested action.

What makes the AI-BI issue card useful

AI-BI combines business facts, business meaning, and knowledge of how the team works.

What AI-BI knowsWhat it means for the Milo team
1. Business facts (live data)What happened?Ho Chi Minh City sell-out is VND 6.18 bn against VND 8.51 bn expected (gap −VND 2.33 bn). Distributor NPP-HCM-01 accounts for 71.8% of the gap.
2. Business meaning (business definitions)How should we read it?The Milo business ontology defines key concepts such as sell-out, expected sales, distributors, SKUs and stock availability, so AI-BI can interpret the figures consistently across time and teams.
3. Ways of working (working context)Who does what?The RSM works with customer service to resolve stock-outs. The brand lead reviews and approves before anything is shared outside the channel.

What AI-BI adds

Live data, business definitions, working context

A chat, a clock and a ledger do not make an analyst. Three things do, and they are what separates AI-BI from a dashboard and from a general AI assistant.

Live data

Today's numbers, read from Nestlé's own data, never changed by the AI.Without it, an AI can only talk in general.

Business definitions

The governed business ontology: what each concept and each number means at Nestlé, approved by a named person. For example: what counts as sell-out, how distribution is measured, which period a share figure covers.Without them, the same question gets two answers.

Working context

How this team works: who owns what, the promotion and season calendar, how each kind of issue is handled, what was asked and decided before.Without it, an answer is one nobody can act on.
  • Live data

    Dashboard
    Yes
    General AI assistant
    No, unless someone pastes it in
    AI-BI
    Yes, read-only
  • Business definitions

    Dashboard
    Built into each report, each function in its own way
    General AI assistant
    None
    AI-BI
    One shared set for every function, approved by a named person
  • Working context

    Dashboard
    None
    General AI assistant
    Only what is typed into the chat
    AI-BI
    The team's own: owners, calendar, ways of handling issues, history
  • Comes to the team

    Dashboard
    No. It waits to be opened
    General AI assistant
    No. It waits to be asked
    AI-BI
    Yes. Every morning, in the group chat
  • Asking why

    Dashboard
    Send a request to an analyst and wait
    General AI assistant
    An answer in general terms
    AI-BI
    An answer with the team's own figures, in the thread
  • Following up

    Dashboard
    Email or chat, no record
    General AI assistant
    None
    AI-BI
    One card per issue, with its size, its cause and a suggested action

Put together, the AI answers with today's figure (live data), means the same thing by it as everyone else (business definitions), and knows who should act and how (working context).

“General AI assistant” means one used on its own, without a connection to the company's data. The comparison describes common patterns; each company may differ. AI-BI does not replace the dashboard: for free exploration across many dimensions, a dashboard is still the better tool.

No new platform

The AI is in your group chat, composed from tools the world already trusts

There is no separate application to open, no new login and no new screen to learn. The AI is a member of the channel, the team's group chat: it speaks first each morning, and it answers when someone @tags it.

Slack · the chat

Where the team already talks. The AI joins the channel like a colleague.

Cron · the clock

Wakes the AI every morning, so the team is told without asking.

Git · the ledger

Every definition and rule is written down, with who approved it. Any change can be undone.

What this means for Nestlé

Nothing to develop
There is no platform of our own to build, roll out or maintain. The effort goes into the business content, not into software.
Nothing new to learn
People already know how to send a message and @tag a colleague. That is the whole interface.
Nothing unproven to trust
Each tool is used every day by teams around the world. The new part is small, and it is where the value is.
Nothing locked in
The idea does not depend on one product. A team on Microsoft Teams keeps Microsoft Teams.
The team stays together
A question asked in the channel is answered in the channel. Knowledge is not lost in private chats with an AI.

Compose, do not build. The chat, the clock and the ledger are solved problems. AI-BI spends its effort on the three things no general tool brings: live data, business definitions and working context.

Rollout plan

Onboard the company once. Then one channel at a time.

AI-BI in one picture: data, definitions, the assistant and where people work, with the block each rollout step builds.
AI-BI in one picture. The numbers name the parts; the orange pills show which block each step of the rollout builds. Slack, Linear and Google Drive are examples; a team keeps the tools it has. Icons are illustrations, not product logos.
  1. Step 0

    Onboard the company

    Build the company-shared business definitions, the governed ontology (block 2 of the picture). Done once and reused by every channel. Kept by the AI Data Governor, the named person who approves definitions.

    2a · shared dictionary2b · definitions repo
  2. Step 1

    One channel at a time

    A channel is whatever the business wants it to be: a whole category such as chocolate, one brand such as Milo, or one team such as Milo's trade team. Each has its own group chat. A slice is how AI-BI reads the business; a channel is where a team works with it, and can cover one slice or several. Build block 3 for it, from 3a to 3c. Kept by the team lead.

    3a · AI assistant3b · schedule (cron)3c · channel repo: operating logic, team knowledge
  3. Step 2

    Monitor, to optimise knowledge and memory

    Watch how the channel is really used, and use it to improve what the AI knows and remembers for that channel. This is the self-improvement loop.

    files in and outmessages between people and the AItasks and issues worked onoutput artifacts and files
  4. Step 3 · optional

    Optimise cost and adoption

    Once a channel is in steady use.

    cost optimisationtuninguser trainingengagement

Then the next channel. Steps 1 and 2 repeat for each new channel; Step 0 is not done again. The first channel is the Milo brand.

Self-improvementEvery interaction between people and the AI provides signals for continuously adapting the AI's ability, knowledge, and memory to better serve the channel.

Use cases

Eight use cases, starting with Milo

The idea covers eight Nestlé Vietnam AI Reinvention use cases across Marketing, Sales and Trade-Operations. Each card opens a demo on synthesized data. The lines below quote or paraphrase the public briefs.

AI-BI Issue Desk — a vision for Nestlé Vietnam AI Reinvention