AI Washing In Business: Separating Vapor From Value (Forbes)

This article was originally written for and published by Forbes.

Everyone's talking AI. Technology vendors claim to have it, but many buyers struggle to separate marketing fluff from real technology that delivers measurable results.

In my conversations with manufacturers, I'm hearing the same refrain: "We know we need AI, but can't afford to deploy 90-180 days of resources to have something fail." The pressure is real and expectations are high, with most setting a 25% productivity improvement as the minimum bar.

The problem is that not all AI is created equal. "AI washing," where providers overstate their AI capabilities, has become serious enough that the SEC started imposing penalties in March 2024. Six months later, the FTC launched “Operation AI Comply,” targeting five companies for deceptive claims, including template libraries marketed as AI solutions.

No business can afford to fall for the hype, least of all manufacturers. When a component supplier fails, or tariffs hit overnight, they need technology that anticipates and acts, not buzzwords in earnings calls.

The AI Washing Problem

Some technology providers announce "AI strategies" with no shipped software. Others bolt a chatbot onto legacy systems and call it "AI-powered" despite doing nothing more than summarizing help docs.

It gets concerning when providers are genuinely trying to add value by integrating large language models into platforms. Platforms can't enforce role-based permissions on AI output, meaning the AI might pull data from records a user shouldn’t access. They lack the audit trails needed for regulated industries to track AI-driven decisions. When permissions change to the underlying system, the AI doesn’t automatically respect those new boundaries, which can create serious security and compliance risks.

Manufacturing software running on 20-plus-year-old, on-premise architectures can't provide the data structure, configurability or real-time connectivity that agentic AI requires. The infrastructure wasn’t designed for AI, and no amount of LLM integration can overcome that fundamental limitation.

Without an evaluation framework, buyers are left guessing which AI claims are legitimate and which are theater.

A Five-Stage AI Maturity Guide For Deciphering Vapor From Value

So, how can you know if AI is vapor or value? This maturity scale, where business value grows exponentially with each step, can help.

Level 1: Press Release AI

In this stage, strategy has been announced, but no software has shipped.

Business value: zero.

Level 2: Summarization And Search

At level two, your chatbot answers questions and surfaces specs, but doesn't take action. Users manually execute every step.

Business value: Minimal. Mostly in the form of time savings by not having to copy and paste questions and documents into ChatGPT.

Level 3: AI-Assisted Decision Making With Limited Action

By now, AI analyzes data and recommends actions, but still requires heavy human oversight. AI suggests. Humans decide and execute.

Business value: Limited. AI identifies patterns and flags issues, but it's employees who spend hours verifying recommendations and manually executing each step.

Level 4: True Agentic Process Automation

Here, the productivity threshold gets crossed. AI agents operate within defined parameters set by business rules. Workflows are completed autonomously and operate on digitized, connected data within a secure, rules-driven engine. Agents make real-time decisions.

This is the level at which manufacturers could achieve that 25%+ productivity gain. In the manufacturing world, it’s where change orders that took three days get handled in hours. Quality complaints requiring manual analysis get triaged and routed automatically.

Business value: Good.

Level 5: Agent-to-Agent Orchestration

Now, multiple AI agents communicate across business functions with no human intervention. Your customer service agent connects with your engineering agent to proactively identify component failures before they become field issues.

This future state is where AI optimizes individual processes and orchestrates entire value chains. A product recall that took days to coordinate across departments happens in hours, with agents automatically pulling serialized asset data, customer records, supplier information and regulatory requirements to execute a comprehensive response. Humans remain in the loop to provide approvals and make decisions, but time-consuming tasks are handled by agents.

Business value: Exceptional.

5 Questions To Ask Your AI Vendor

In my experience, most manufacturing tech providers sit between levels one and three. Many can't scale without replatforming.

To ensure success, ask vendors:

1. Production Proof: Can you show AI running in production with real data? Ask for a live demo with real workflows and measurable results.
2. Security And Compliance: How does AI enforce role-based permissions and maintain audit trails (for regulated industries)? AI must inherit your existing security model, log every AI action with user attribution and produce compliance reports showing who accesses data and when.
3. Governance: What business-rules engine governs autonomous actions? Vendors should explain specific guardrails and approval thresholds, and demo how AI operates within your existing business process rules.
4. Integration: For manufacturers, can agents connect data across business systems like PLM, CAD, ERP, quality and supply chain? AI agents should pull data from multiple connected systems in real time to provide comprehensive answers.
5. Architecture: Is your architecture cloud-native or AI retrofitted onto legacy infrastructure? Best-in-class infrastructure is built from the ground up with cloud-based platforms and APIs, not on an on-premise system with AI bolted on. Retrofitted AI can’t adapt when business rules or permissions change.

Level one and two providers talk future roadmaps. Levels four and five show production deployments with measurable outcomes.

Final Thought

Real AI delivers measurable outcomes tied to faster time-to-market, reduced errors and healthier margins. Today, hesitation comes at a high cost. Companies that fall for AI washing waste resources. Businesses that deploy genuine agentic AI gain a competitive advantage that scales.

The choice is yours. But first, you need to know how to tell the difference.

Want to see what real agentic AI actually do? Get a demo today.