Copy article

Is Cloud AI the Concorde of the Modern Age?

ended 28. May 2026

Microsoft and Uber have both exposed the same awkward truth about AI tokenomics: sometimes the tool works, and the business problem morphs.

Concorde was a flashy, expensive answer to something humans could already do. The selling point was speed. A Boeing 747 was slower, less glamorous and far more useful at scale. Concorde was retired in 2003 because prestige couldn’t outrun economics.

Cloud AI may be reaching its own Concorde moment. Business automation was not invented in 2022. Microsoft Office macros, scripts, robotic process automation and no-code tools such as Zapier have handled predictable work for years.

Old-school deterministic automation says: if this happens, do that. Generative AI says: give the machine a goal, feed it context, pay per token, then supervise whatever comes back. That functionality provides flexibility, but it is not automatically cheaper, safer or more insurable. For SMEs, the awkward question is not “Should we use AI?” It is: “Which jobs need probabilistic interpretation, and which ones just need boring workflow logic?”

That distinction matters because the hidden costs are now visible. Token pricing can turn a pilot into a budget leak. Agentic coding tools can burn through usage in ways finance teams struggle to forecast. Professional indemnity and cyber insurance may become harder to rely on where industry regulations, client data, opaque AI decisions and vibe-coded internal tools are mixed together without proper controls.

Concorde proved that speed alone is not a business model. Cloud AI may prove the same thing in software.

We’d like your views:

  • If more companies had properly implemented deterministic no-code automation earlier, how much of today’s rush to LLMs would be unnecessary?
  • Where is the sensible boundary between predictable workflow automation and AI agents that need supervision, logging and professional accountability?
  • If a consultant routes client data through Claude Code on a consumer Pro or Max account, should insurers treat that as ordinary tool use or a breach of professional controls?
  • At what point does per-token API pricing destroy the ROI case, once retries, context windows, agent loops and human checking are factored in?
  • What happens to firms that rebuild operations around AI workflows and then discover the token bill is too rich?

 

3 responses from the Newspage community

Copy all

Star Quote
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
Big businesses have been automating with macros, VBA, formulas, inbox rules, scripts and SaaS for years. SMEs often had less appetite for that kind of process work. At their scale, they could live with the duct tape and clunkiness.

Skipping this step left many without solid processes, tested edge cases, or clean, consistent data. No-code platforms were ready to formalise what the duct tape was holding together, but the nettle was never grasped.

Then gen AI arrived, and SMEs grabbed a probabilistic tool without the foundations a logic-based one would have forced them to build. Pain follows as muddy data feeds a system that guesses. Processes fall apart at the edges because nobody mapped them. Token bills climb while teams run through treacle salvaging the output.

Cloud AI is convenient, but local AI offers small firms something better: cheaper tokens, better privacy, and less risk of a supplier changing the price, quality or rules overnight. Get the benefits with fewer costs.
Copy

Editor
Katrina Young
Futurist | Chief Technology Officer CTO/CDO Ethical AI & Digital Transformation Strategist at KYC Digital
Microsoft and Uber’s recent admissions expose the same problem: a tool can work while the underlying business problem gets worse. That is the part of the Concorde analogy worth keeping. The real issue is not whether cloud AI is “good” or “bad.” It is whether organisations are routing the wrong type of work through probabilistic systems. Deterministic automation handles predictable workflows at known cost. Generative AI handles ambiguity, interpretation and unstructured input. The mistake is treating every workflow as an interpretation problem when much of it is just logic, then paying per token for the privilege. The hidden risk is not token pricing. It is accountability drift. Once client data, AI outputs and internal decisions move through tools not configured for enterprise-grade traceability, logging or governance, many firms can no longer explain what happened, why it happened, or who approved it. The architecture has to come before the bill arrives.
Copy

Editor
Nouran Moustafa
Practice Principal & IFA at Roxton Wealth
The mistake many firms are making is treating AI as the answer before defining the job. A lot of business processes do not need intelligence; they need discipline. If a task is predictable, rules-based and repeatable, traditional automation may be cheaper, safer and easier to audit than an AI agent burning tokens in the background.

AI is powerful where judgement, language, context or interpretation is genuinely needed. But using it for basic workflow logic is like hiring a consultant to press a light switch.

For SMEs, the real question is not “how do we use AI?” It is “which parts of the business justify AI risk and cost?” Token bills, data controls, supervision, audit trails and insurance all matter. A clever pilot can quickly become an expensive liability if nobody understands how it scales.

The winners will not be the firms using the most AI. They will be the firms boring enough to automate what is predictable and brave enough to use AI only where it genuinely adds value.