The AI Solution Mix — How the Six Categories Work Together in a Real Enterprise
By Diego Navia · BizBlocz · May 2026
The closing article of the AI Explained series. Start with the overview →
Where the Series Leaves Us
Over the last seven articles, this series covered each of the BizBlocz AI Six — the six categories of AI in common enterprise use:
- Machine Learning — it predicts
- Agentic AI — it acts
- Generative AI — it creates
- Natural Language Processing — it understands
- Document AI — it reads paperwork
- Computer Vision — it sees
Each article covered what the category is, how it works, where it fits, and where another category leads instead. Taken together, the six cover the landscape.
But knowing the six isn't enough. Enterprise processes don't run on one of them. They run on a mix — and the mix is different for almost every process in a company.
This article covers how that mix actually looks in practice, why it changes from one subprocess to the next, and how a process-level AI strategy differs from a technology-level AI strategy.
The Scale of the Selection Problem
There are more than 46,000 AI startups globally and more than 8,000 vendors competing for enterprise attention (Stanford HAI, 2025 AI Index Report; AIMultiple, Enterprise AI Company Landscape, 2026). Across the six categories covered in this series, the vendor landscape is dense enough that even experienced transformation leaders struggle to map it to their operations.
The result is predictable. MIT NANDA's State of AI in Business 2025 found that 95% of enterprise generative AI pilots delivered no measurable P&L impact — despite an estimated $30–40 billion in global enterprise generative AI investment. The technology worked. The selection didn't.
Gartner's 2025 Hype Cycle captured the same dynamic from a different angle: generative AI has entered the Trough of Disillusionment, while agentic AI sits at the Peak of Inflated Expectations. The categories doing the largest share of enterprise work — machine learning, agentic AI, document AI — rarely headline the keynotes.
Worldwide generative AI spending reached $644 billion in 2025 (Gartner, March 2025). Enterprise generative AI application spending alone grew from $1.7 billion to $37 billion since 2023, capturing 6% of the global SaaS market (Menlo Ventures, State of Generative AI in the Enterprise, 2025).
That is the asymmetry this series has been pointing at. Generative AI gets roughly 60–70% of enterprise AI budget today. Across 127 enterprise subprocesses, it accounts for 18% of aggregate AI value. The gap between those two numbers is where most enterprise AI investment is being wasted.
What the Aggregate Mix Looks Like
Mapping the six categories across a reference taxonomy of 127 universal business subprocesses — drawing on 245+ data points across 30+ independent research publications — the weighted portfolio looks like this:
| AI Category | Aggregate Portfolio Share |
|---|---|
| Machine Learning | 32% |
| Agentic AI | 22% |
| Generative AI | 18% |
| Natural Language Processing | 14% |
| Document AI | 9% |
| Computer Vision | 5% |
Machine learning leads, but no single category dominates. Generative AI is a real 18% — not marginal, but not the 60–70% of budget it absorbs in most organizations. Document AI and computer vision are small shares of the aggregate but foundational where they apply (finance and operations for document AI; manufacturing, logistics, and healthcare for computer vision).
The aggregate view is useful for calibration. It is not how an AI strategy gets built. The real unit of analysis is the subprocess.
Same Company, Different Process, Completely Different Mix
The mix doesn't just vary across industries. It varies across subprocesses within the same function, in the same company, on the same floor. Here are four subprocesses that might all sit inside one CFO's organization.
Invoice Processing (AP01)
| Category | Share |
|---|---|
| Document AI | 40% |
| Agentic AI | 30% |
| Machine Learning | 10% |
| Generative AI | 10% |
| Computer Vision | 5% |
| NLP | 5% |
This is a document-first process. The entry point is a piece of paper or a PDF, and everything downstream depends on how accurately the data is extracted from it. A generative AI chatbot deployed as the primary solution is addressing roughly 10% of the opportunity with most of the budget.
Payment Execution (AP02)
| Category | Share |
|---|---|
| Agentic AI | 50% |
| Machine Learning | 20% |
| NLP | 15% |
| Generative AI | 10% |
| Document AI | 5% |
Same department. One subprocess downstream from invoice processing. The dominant category flips entirely. Payment execution is a rules-based, high-volume transaction workflow — selecting invoices, applying discount terms, triggering bank transfers, processing exceptions. The category that matters here is agentic AI.
Demand Forecasting (DP01)
| Category | Share |
|---|---|
| Machine Learning | 65% |
| Generative AI | 15% |
| NLP | 15% |
| Agentic AI | 5% |
A prediction-first process. The value comes from an algorithm's ability to detect demand patterns across hundreds of SKUs, channels, and time periods. Generative AI can help narrate the forecast. It cannot produce it.
Automated Defect Detection (QI01)
| Category | Share |
|---|---|
| Computer Vision | 65% |
| Machine Learning | 20% |
| NLP | 10% |
| Agentic AI | 5% |
The dominant category is computer vision — cameras and image analysis models inspecting products on a production line at speeds no human inspector can match. A language model has no primary role in this process.
Four subprocesses, four completely different technology profiles. An "AI for Finance" or "AI for Operations" strategy that doesn't operate at this level of specificity is a budget allocation, not a strategy.
The Handoffs Between Subprocesses
Subprocesses don't run in isolation. Invoice processing finishes extracting the data. Payment execution needs that data to execute. Between them is a handoff that crosses systems — the OCR tool, the ERP, the approval workflow, the payment gateway. None of these were built to talk to each other natively.
That gap exists at every subprocess boundary in every enterprise. The average enterprise runs over 900 applications, and only 29% are integrated (MuleSoft, Connectivity Benchmark Report, 2025). The rest rely on people — or increasingly, on automation — to bridge the gaps.
It is unglamorous work. Automated at scale, it delivers real labor savings and real cycle-time reductions. It doesn't show up in any single subprocess's AI mix, because it lives between them, not within them. It is one of the practical reasons agentic AI is projected to rise from under 5% of enterprise applications in 2025 to 40% by 2026 (Gartner, "Gartner Predicts 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026," August 2025).
Why Enterprises Get the Selection Wrong
Three root causes show up consistently.
1. Vendor incentives are not aligned with process needs. Every AI vendor leads with its own category. The document AI vendor says AP needs intelligent document processing. The agentic AI vendor says AP needs agents. The generative AI vendor says AP needs a copilot. All three are partially right. None has an incentive to say what percentage of the problem their category actually addresses.
2. The strategy conversation happens at the wrong altitude. "AI for Finance" is a department label, not a process specification. Finance alone typically contains 11 processes and 22 subprocesses, each with a different mix. Decisions made at the department level are almost guaranteed to over-invest in some subprocesses and under-invest in others.
3. The decision framework has been missing. Gartner, Forrester, McKinsey, BCG, and Deloitte have all published valuable AI strategy research — on vendor landscapes, use-case prioritization, and value gaps. Before BizBlocz published the 127-subprocess mix, none had published a framework that prescribed which AI category should lead for each specific subprocess with disclosed methodology and source-level transparency. The gap was structural.
That structural gap is part of why 95% of generative AI pilots deliver no measurable impact, why 30% of generative AI projects are abandoned after proof of concept (Gartner, "Gartner Predicts 30% of GenAI Projects Will Be Abandoned After Proof of Concept by End of 2025," July 2024), and why only 20% of organizations report generating revenue growth from AI despite 74% expecting to (Deloitte, State of AI in the Enterprise 2026).
The BizBlocz VALUE Assessment
Across the 127 subprocesses in the BizBlocz taxonomy, the VALUE assessment is built to answer four questions for each one:
- What is the savings range — directional financial opportunity from applying AI to this subprocess, expressed as a range rather than a point estimate
- What is the confidence level — how strong the evidence is for that range, based on how many independent research sources converge and how closely their findings align
- What is the AI solution mix — the proportional breakdown across the six categories, showing which should lead and which play supporting roles
- What are the sources — the specific research publications and data points that inform each allocation, with link-outs
The assessment takes roughly 10 minutes. It runs against the 127-subprocess taxonomy, with a waterfall confidence methodology layered over 245+ data points from 30+ research publications including Everest Group, McKinsey, Gartner, Deloitte, BCG, Forrester, MIT, Stanford HAI, and O'Reilly. Mix percentages are directional — calibrated estimates grounded in the research, not decimal-precise prescriptions — and the methodology is disclosed in full.
The practical impact is immediate:
- Before a vendor conversation: know which category should dominate for the process in question. If a vendor is selling generative AI for a process that is 65% machine learning, the right question is which percentage of the problem their solution actually addresses.
- During pilot scoping: size the pilot to the right category. A document AI pilot for invoice processing is a fundamentally different scope than a generative AI pilot for invoice processing, and the return profile is different.
- During budget allocation: distribute investment across categories in proportion to where the evidence says the value lives, rather than where the vendor pipeline is loudest.
- Post-implementation: measure against the right baseline. A process that is 40% document AI, where only the agentic AI component was deployed, has addressed a minority of the opportunity — and that should be visible in the result.
The tool is free to run, takes no signup, and produces a full subprocess-level output: bizblocz.com/ai-diagnostic →
The Question That Matters
"Should we use AI?" is a settled question.
"Which vendor?" is a premature question.
The useful one is the same one that opened this series, now with seven articles of context behind it:
Which AI category, applied to which specific process, in what proportion, backed by what evidence?
That question separates the roughly 5% of companies achieving AI value at scale (BCG, "Are You Generating Value from AI? The Widening Gap," September 2025) from the majority still running pilots that go nowhere.
The map exists. Each of the BizBlocz AI Six has its role. Most processes use a mix of several. Every allocation in the 127-subprocess taxonomy has a rationale, and every rationale has a source.
That is where a process-level AI strategy starts.
The Series in One Page
| Article | Category | Aggregate Share | One-line Summary |
|---|---|---|---|
| AI-2 | Generative AI | 18% | Creates content. Fits where the deliverable is language. |
| [AI-3] | Machine Learning | 32% | Predicts outcomes. The largest single share of enterprise AI value. |
| [AI-4] | Agentic AI | 22% | Takes action. The fastest-growing category on the 2025 Hype Cycle. |
| [AI-5] | Computer Vision | 5% | Interprets visual input. Foundational in manufacturing, logistics, healthcare. |
| [AI-6] | NLP | 14% | Understands language. Often invisible inside platform features. |
| [AI-7] | Document AI | 9% | Reads paperwork. The category with the deepest roots (OCR since 1959). |
Six categories. One mix per process. 127 subprocesses mapped.
Transform smarter. Intelligence you can build on. bizblocz.com
Sources
- Stanford HAI, 2025 AI Index Report — 46,200+ AI startups globally
- AIMultiple, Enterprise AI Company Landscape (2026) — 8,000+ AI vendors
- MIT NANDA, State of AI in Business 2025: The GenAI Divide — 95% of GenAI pilots deliver no P&L impact
- Gartner (March 2025) — Worldwide GenAI spending: $644 billion in 2025
- Gartner, Hype Cycle for Artificial Intelligence 2025 — GenAI in Trough of Disillusionment; Agentic AI at Peak of Inflated Expectations
- Gartner (July 2024) — 30% of GenAI projects abandoned after proof of concept by end of 2025
- Gartner (August 2025) — 40% of enterprise apps will feature task-specific AI agents by 2026
- Menlo Ventures, State of Generative AI in the Enterprise (2025) — $1.7B to $37B enterprise GenAI application spending since 2023
- BCG, Are You Generating Value from AI? The Widening Gap (September 2025) — Top 5% achieve value at scale
- Deloitte, State of AI in the Enterprise 2026 — Only 20% generating revenue growth from AI; 74% hoping to
- McKinsey, The State of AI (November 2025) — 88% regular AI use; 80%+ see no EBIT impact from GenAI
- MuleSoft, Connectivity Benchmark Report (2025) — Average enterprise runs 900+ applications; only 29% integrated
- Everest Group, IDP PEAK Matrix (2024) — Document AI dominance in invoice capture
- Grand View Research, OCR Market Report (2025) — $17B OCR market
- BizBlocz AI Mix Rationale — 127 subprocess-level AI technology allocations with source attribution across 6 primary research sources
