The $15 Billion Rebrand: Why Our Revolutionary AI is 25 Years Old

I attended a Genesys partner conference recently. The presentation deck and speakers were impressive.

They talked about their platform evolution, moving beyond CCaaS to become a full orchestration platform. They dropped every AI buzzword: A2A (agent-to-agent), MCP (Model Context Protocol), agentic workflows, AI Agents, multi-modal AI experiences.

Then they shared some real-world use cases.

Transcription feeding AI-generated wrap-up summaries. Agent assistant suggesting responses. A chatbot.

That was it. That’s what customers actually had working in production.

But here’s my biggest takeaway: It wasn’t about their AI capabilities at all. It was about how much they need partners to support AI implementation. Translation? Customers can’t figure this out on their own, and Genesys can’t scale the support required to get people live.

And it’s not just Genesys. On February 23, 2026, OpenAI, the company that arguably started the current AI hype cycle, announced multi-year partnerships with BCG, McKinsey, Accenture, and Capgemini. Why? Because enterprises can’t implement their AI without help from four of the world’s largest consulting firms.

They even said the normally quiet part out loud: “The limiting factor for seeing value from AI in enterprises isn’t model intelligence, it’s how agents are built and run in their organisations.“

Read that again. The most prominent AI company on the planet is admitting that AI itself isn’t the hard part, everything else is.

Major software vendors have poured over $15 billion into AI acquisitions and initiatives since 2022. ServiceNow spent $2.85 billion on Moveworks. NICE dropped $955 million on Cognigy. Salesforce built a billion-dollar AI venture fund. Genesys received $1.5 billion in fresh funding from ServiceNow and Salesforce.

Want to know what customers are actually getting?

Call summarisation and wrap-up. Email summarisation. Basic sentiment analysis.

The same features that were technically possible when Windows 95 launched.

What’s Really Happening

If you’ve been in this space for more than five years, you’ve seen this movie before. Different actors, same script, bigger budget.

The data tells the brutal truth. Despite $15 billion in investment, only 3-6% of gen AI operations are actually scaling. Gartner found that by the end of 2025, at least 50% of generative AI projects were abandoned due to poor data quality, inadequate risk controls, escalating costs or unclear business value.

So where’s all that money going?

The Great AI Rebrand

Let’s talk about what’s being called “revolutionary AI capabilities.”

Call summarisation: Genesys first launched this in June 2023. Automatic Text Summarisation (ATS) was actually invented back in 1958. Then in 1996 we saw the birth of multi-lingual extraction, and by 2015 neural networks and deep learning began to revolutionise the field.

Sentiment analysis: Zendesk charges approximately $50/agent/month extra for this. Originally referred to as Opinion Mining, has been around since the 1960s. Current AI systems achieve 75-85% accuracy. Human-level detection? 80-85%. Yes, AI-powered provides a significant advantage when it comes to speed and scale. You’re still being charged premium prices ~20-year-old tech.

Agent assist: Real-time guidance during calls sounds impressive until you realize knowledge base search and agent guidance systems existed in the 1990s. The “AI” part mostly means faster search. Same capability, new label, higher price.

AI adds speed and scale, but this isn’t true innovation. It’s mostly AI rebranding.

The Demo-to-Production Gap

Here’s what really happens behind the scenes in vendor demos: carefully selected scenarios, hand-picked data, perfectly engineered prompts. Everything works beautifully.

In production: the AI lacks context for multi-part questions, hallucinates confident answers that are completely wrong, and requires constant human intervention.

But the real issue isn’t the technology at all. It’s the gap between what’s possible and what’s actually required to implement.

The promise: Revolutionary AI that transforms your contact centre. Simple to implement. Immediate ROI.

The reality: Customers need to validate costs, prove the system works in their specific environment, customise for their use cases, test at small scale before rolling out. That generally requires consultants and time.

When your “plug and play AI revolution” requires an army of implementation partners to customise, validate, and prove value before you can scale, did you actually receive what was promised?

Sounds more like you were sold a dream and the beginning of a lengthy services contract.

Air Canada learned this the expensive way. Their chatbot provided incorrect bereavement fare information. When the airline argued the chatbot was “a separate legal entity responsible for its own actions,” a tribunal found them liable for negligent misrepresentation. The chatbot is no longer on their website!

One hospital’s AI scheduling system “worked beautifully for about six weeks.” Then patients were misrouted, slots double-booked, calls got longer. The demo was theater. Production was chaos.

And finally, the UK-based start-up Builder.ai who claimed it’s “advanced code-writing AI” could build apps faster and cheaper than traditional methods. Turns out their AI was actually 700+ human developers in India. AI theater at its very best!

The Pricing Structure That Tells the Story

Zendesk’s pricing model reveals exactly how vendors extract value from AI hype.

Base Suite Team Plan: $19/agent/month. Sounds reasonable.

But to access AI features, you need Professional ($$) or Enterprise ($$$), then add Advanced AI (+$$/agent), AI Agents (+$$/agent), Quality Assurance (+$$/agent), and Workforce Management (+$$/agent).

Real-world costs hit 2-3x the base rate after all the ‘essential’ add-ons.

Salesforce goes bigger. Base Enterprise Edition costs $165/user/month. Add Agentforce ($$$), Einstein Conversation Insights ($$), and Revenue Intelligence ($$$). You’re now at in excess of $500/user/month.

Luxury prices for dial-up era features. At scale, these costs are absurd.

And that’s before you factor in the consulting fees. OpenAI’s new Frontier platform doesn’t come with a plug-and-play promise. It comes with a McKinsey engagement, an Accenture delivery team, and dedicated practice groups who need to be certified before they can even start. If the world’s best-resourced AI company needs that ecosystem just to get its product into production, your vendor’s “seamless implementation” slide deserves a very hard look.

What about the Analysts? Surely they’re calling this out.

Gartner placed GenAI in the “Trough of Disillusionment” in their 2025 Hype Cycle. Translation: reality is setting in.

Forrester’s July 2025 report didn’t pull punches: “Despite the hype, many organizations are discovering that AI alone isn’t delivering the transformative results they expected.” Their assessment of current deployments? “Simple and safe tools that improve how we do the same things we have always done—not actually changing service.”

McKinsey found that despite 80% of respondents investing in generative AI, only 3% have scaled a GenAI use case in operations. Of 876 companies surveyed, only 46 attribute more than 10% of Earnings Before Interest and Taxes (EBIT) to generative AI.

Gartner predicts that over 40% of agentic AI projects will be canceled by 2027 due to inflated claims and underwhelming outcomes.

Here’s what those analyst reports aren’t saying directly: The vendors paying for Gartner Magic Quadrants and Forrester Waves are the same ones burning billions on AI. The reality? That “revolutionary AI platform” has a 40% chance of being binned before 2027.

If you need one more signal that the gap between AI promise and production reality is real, look no further than OpenAI itself. Not a failed case study. Not a sceptical analyst. The company. They just hired McKinsey, BCG, Accenture, and Capgemini to help their enterprise customers bridge it.

Three Questions Every Leader Should Ask

Before you sign that seven-figure contract for “revolutionary AI capabilities,” ask these questions:

1. What problem are we actually solving? If the answer is “we need AI because everyone has AI,” please stop. Technology without a specific business problem is just expensive theater.

2. Can you show me this working in production, not a demo? Insist on customer references running similar scale and complexity. Talk to their implementation teams, not just executives who signed the deal.

3. What’s the real total cost of ownership? Add up all the “optional” modules that turn out to be essential. Include implementation, integration, testing, ongoing support, and the inevitable customisation. Then double it, because you’re bound to hit edge cases.

What This Means for You

The organisations succeeding with AI aren’t buying the biggest platforms or the flashiest features. They’re identifying specific, measurable real problems, and implementing focused solutions.

They have real problems to solve and are using AI as a tool, not a magical fix all.

They’re measuring actual outcomes, not made up “AI adoption metrics.”

They’re honest about what AI can and can’t do … yet.

And they’re not paying premium prices for 25-year-old technology wrapped in new AI-powered marketing.

If you’re evaluating AI right now, you have leverage. Vendors are desperate to show revenue growth to justify those billion-dollar acquisitions. They’ll negotiate.

But more importantly, you have permission to be sceptical. When a vendor shows you call summarisation and calls it revolutionary AI, you can point out that technology is older than Google.

When they promise 80% automation, you can ask for the customer references who actually achieved it. (Spoiler: there’re won’t be many.)

The emperor has no clothes. And now even the tailor is admitting it.

The question isn’t whether AI will transform contact centres and business operations. It will, eventually.

The question is whether you’re going to pay transformational prices and costly services engagements for incremental features while everyone figures out how to implement it.