Area

Generative AI

Generative AI, in enterprise use, means applications built on large language models: drafting, summarising, extracting and classifying. The distinction from agentic AI is that a generative application answers what it was asked, rather than deciding what to do next.

Generative AI

Wipro

Wipro launched ai360 in 2023 with a $1 billion, three-year AI investment commitment, backed by the…

Generative AI

EY

EY launched its EY.ai Agentic Platform in March 2025 with NVIDIA, building custom frameworks for designing…

Agentic AI

Cognizant

Cognizant positions itself as an AI Builder and delivered more than $21 billion in 2025 revenue,…

Agentic AI

IBM Consulting

IBM Consulting builds its generative AI practice on watsonx, a platform Gartner named a Leader across…

Generative AI

BCG X

BCG X, the build-focused arm of Boston Consulting Group, creates industry-specific AI platforms such as Auto…

AI Consulting

McKinsey & Company

McKinsey's QuantumBlack unit began as an independent Formula 1 racing analytics firm before McKinsey acquired it…

Agentic AI

Deloitte

Deloitte built Zora AI, an agentic product platform shipping ready-to-deploy AI agents that perceive, reason, and…

AI Consulting

Bain & Company

Bain and OpenAI have run a joint services alliance since 2022, and Bain became an OpenAI…

Agentic AI

Anthropic

Anthropic built its position by winning the enterprise market that OpenAI built its reputation on. The…

About Generative AI

Generative AI, in enterprise use, means applications built on large language models: drafting, summarising, extracting and classifying. The distinction from agentic AI is that a generative application answers what it was asked, rather than deciding what to do next.

At a glance
Firms tracked
21
Typical buyer
Function leads
Common first use
Document extraction
Typical pilot length
4 to 8 weeks
Main failure mode
No accuracy baseline
Maturity
Mainstream

Area reviewed 12 Sep 2026

What the work involves

Mostly plumbing, not prompting

The model is rarely the hard part. The work is retrieval over your own documents, output validation, human review paths and the accuracy measurement that tells you whether any of it helped. Firms that lead with prompt engineering as a discipline are describing the easiest 10% of the project.

Who buys it

Function leads with a document backlog

Legal, claims, procurement and finance teams buy this first, because they hold large volumes of text that people currently read manually. The budget usually sits in the function rather than in IT, which speeds up the pilot and slows down the production deployment.

Firms in this area

A crowded field, 42 firms tracked

M

Marlowe AI

Document-heavy LLM applications, regulated industries
Bengaluru, 2022
Strongest here


A

Aulric Labs

Generative work adjacent to its agent platform
London, 2021
Also covers


V

Verrow Data

Retrieval pipelines and evaluation tooling
Dublin, 2017
Infrastructure


A

Anthropic

Model provider, not a delivery firm
San Francisco, 2021
Model

See all 42 firms in this area

What to ask

Four questions worth more than a demo

  1. What is our current accuracy on this task, measured how? Without a baseline there is no result.
  2. What happens to an output the model is unsure about, and who sees it?
  3. Where does retrieval come from, and who maintains that index after go-live?
  4. What is the review sample rate in production, and does it shrink over time or stay fixed?

Research notes

How this area was assessed

This is the largest and least differentiated area we cover. We track 42 firms and rank none of them overall, because the work is too task-specific for a single ordering to be useful. Instead we assess by use case: document extraction, drafting, classification and summarisation are effectively four different markets with different leaders.

Use case Firms with production references Assessment status
Document extraction 14 Ranking published Q4
Drafting and summarisation 19 In progress
Classification and routing 11 In progress
Conversational support 9 Not started

Sources

  1. Firm interviews, 42 firms, March to September 2026
  2. Client reference interviews, 23 completed
  3. Buyer interviews on accuracy baselines, 2026
© 2026 Top AI Firms. Independent research \u{2014} we take no payment for placement.How we assess a firm · About ·