ATPL

AI · Start right

Know where AI pays off before you spend.

For owners and directors under pressure to "do something with AI": two to three weeks inside your operation, ending in a ranked list of where it pays, what it costs, and what to leave alone.

Describe the problemAll AI capabilities

Advise or build

Where it fits, where it doesn’t

Start here if you are not sure AI is the right spend. The map comes before the model.

  1. Fits · 01

    Businesses with real operations and real data who want a plan before a pilot.

  2. Fits · 02

    Leadership willing to give access to the people doing the work, not only to managers.

  3. Fits · 03

    Companies that want the "no" list as much as the "yes" list.

  4. Skip · 01

    A business that has already decided what to build; go straight to a proof of concept.

  5. Skip · 02

    A request for a strategy deck for investors; we write plans that someone will execute.

  6. Skip · 03

    Companies with no digital records at all; the first step is systems, not AI.

Your data

During the study we read your data where it lives, under access you grant and can revoke, and take away only the aggregated figures and samples the report needs. Nothing is uploaded to a model provider. The report and the opportunity register are yours.

Before the build

What to decide before you buy a model

A short, honest study of where AI pays off in your business, and where it will not.

  • 01 · Hype

    Every vendor says AI; none of them has asked how your business actually works.

  • 02 · Where

    You cannot tell a real opportunity from a demo that will cost a year.

  • 03 · Ready

    Your data may not be ready, and nobody has told you that plainly.

A clear map

Where AI pays, and where it does not

  • 01

    Operation walk-through

    Days spent with the people who do the work: billing, dispatch, accounts, support. We watch, we count hours, we read the data.

  • 02

    Opportunity register

    Every candidate use, with the hours or rupees at stake, the data it needs, the risk if it is wrong, and whether a rule would do instead of a model.

  • 03

    Data readiness check

    For each top candidate, a plain statement of whether the data exists, where, how clean it is, and what fixing it would take.

  • 04

    Ranked plan with costs

    The first three projects in order, each with a build estimate, a running cost estimate, the measure of success and the person who should own it.

  • 05

    A written "not now" list

    The ideas we looked at and recommend against, with reasons, so the question does not come back next quarter.

How it ships

Jobs first, models second

  1. map/01

    Understand

    We start with your numbers, not with AI: where the hours go, where the errors come from, where cash is stuck.

  2. map/02

    Design

    Candidates are scored on hours saved, data available and risk, and we argue each one with you; some drop out because a rule or a report would do.

  3. map/03

    Build

    The outcome is a document and a working session, not a slide deck: ranked projects, costs, owners and the first proof of concept scoped.

  4. map/04

    Stay

    We revisit the plan after the first project ships, because what you learn there usually re-orders the rest.

Problem to brief

A vague ask becomes a decision brief

Where AI pays, where it does not, and what to prove first — advice you can act on.

What a client typed

I run a mid-sized garment export house with about two hundred people and our board keeps asking what we are doing about AI, but honestly I do not know where it would help, and I do not want to waste a year on a pilot.

Brief ready

What it becomes

Problem
Board pressure to adopt AI with no clear view of where it would pay.
Scale
Around two hundred people across merchandising, production, documentation and accounts.
Systems
An ERP, Tally, export documentation templates, buyer portals, email.
First step
A three-week study ending in a ranked list of three projects with costs, and a named list of ideas to leave alone.

How we work

ATPL has built and run its own products, Ethos since 2013 among them, so its advice on AI comes from operating systems rather than selling them. ATPL will confirm which advisory engagements may be named.

    Chosen for the job

    Python · Node.js

    Decide the shape

    Before you start

    Plain answers on scope, data and what a first proof usually takes.

    An AI consultant finds where AI would save hours or money in your specific operation, checks whether the data to do it exists, estimates the cost, and tells you which ideas to leave alone. ATPL does this in two to three weeks on site and in your data, and delivers a ranked plan rather than a deck.

    If the work you want to automate is recorded in a system, repeated often, and has a measurable outcome, you are probably ready for one project. If the records are on paper or in heads, start with systems. ATPL’s data readiness check answers this candidate by candidate.

    An ATPL AI study is a fixed-price engagement of two to three weeks, priced on the size of the operation and the number of departments covered. The output is a ranked project list with build and running cost estimates, so the next spend is a decision, not a guess.

    Hire a team to do the first project with your people in the room, then decide. ATPL works alongside your staff so the knowledge stays, hands over code and documentation, and tells you when something is simple enough to keep in-house.

    Continue to