Skip to content
Abhishek_Garg

AI systems

More time for judgment. Less time on repetition.

How I actually use AI in marketing work: as an execution layer for research, outreach, and prototyping. Everything below is how I work day to day, not a product claim.

The analytical side

Better questions.
Sharper decisions.

Three ways to turn complexity into a decision. Explore the models.

Find the shape of demand.

A straight line is a starting point. Compare it with a curve to see where the relationship changes.

Interactive illustration · synthetic data

Model form
00252550507575100100Response indexMarketing input index
0.93 R² on illustrative data26 observations · least squares

The curved fit captures the flattening in this sample. Better fit alone does not establish causality or out-of-sample accuracy.

Methods & project context

The conjoint interaction connects to my independent Conjoint Research Studio project: study design, conditional multinomial logit, ratings-based ridge regression, and scenario simulation. These three visual examples use synthetic inputs, not client results.

Method references: Least-squares regression · Conjoint utilities · Marketing mix modeling

Independent tool

Conjoint Research
Studio.

From study design to product trade-offs.

Design

An offline desktop workflow for preparing studies and importing choice or ratings responses.

Estimate

Conditional multinomial logit for choices and ridge regression for ratings, with part-worth and attribute-importance views.

Explore

Product scenarios, preference-share simulation, and sensitivity analysis. PDF and spreadsheet exports for sharing findings.

Workflows

01

Research → insight

Perplexity and Claude compress the gathering half of market and competitor research, so the time goes into interpretation and the argument I actually want to make.

  • Perplexity
  • Claude
  • ChatGPT
  1. Question
  2. Source sweep
  3. Synthesis
  4. Point of view
02

Signals → outreach

Account research and drafting are AI-assisted; the angle, the offer, and the final wording are mine. Used for the structured outbound programme at Satsheel.

  • Claude
  • ChatGPT
  1. Account signal
  2. Context
  3. Drafted angle
  4. Human edit
03

Idea → prototype

Codex and Claude let me put a landing page, a calculator, or a reporting view in front of people while the idea is still worth arguing about.

  • Codex
  • Claude
  1. Sketch
  2. Prompt
  3. Working page
  4. Test

Principles

Leverage, not theatre
The useful uses are narrow: outbound personalisation, research acceleration, faster drafts, and small internal tools. Anything else is decoration.
The judgment stays human
AI shortens the distance to a first version. Deciding what a business should do with it is still the part that has to be earned.
Prototype before you argue
A rough working page settles a debate faster than a deck. Rapid prototyping earns its keep when it makes an idea testable.
Scattered data points converging into a single decision point, then rising as one growth curve.

Tools in active use

  • Claude

    Long-form reasoning, drafting, and building

  • Codex

    Prototyping pages, scripts, and internal tooling

  • ChatGPT

    Fast iteration on copy, structure, and analysis

  • Perplexity

    Sourced market and competitor research