Reality Check: Where Generative AI Still Falls Short for Business
We’ve talked before about the difference between AI hype and real progress. But even as awareness grows, the buzz around generative AI continues to surge, often with little reference to its practical limitations.
For small businesses, heritage organisations, and applied technology developers like us, it’s time to take a closer look at where generative AI actually delivers, and where it still falls short.
This isn’t about dismissing the technology. It’s about setting realistic expectations, identifying use cases that genuinely benefit from generative tools, and flagging those that need more thought, context, or control.
1. Outputs Without Ownership
Generative models are great at producing fast, fluent content, but that doesn’t mean the content is reliable, original, or legally usable.
Can you copyright it? Often, no.
Can you trust it? Not without human oversight.
Can you use it commercially? That depends, especially when it’s trained on questionable datasets.
For SMEs or organisations in creative sectors, this creates uncertainty. If your output includes AI-generated design, copy, or imagery, you still need clear processes for checking, editing, and (where needed) disclaiming that work.
Without that layer, businesses risk reputational or legal consequences, especially as copyright legislation evolves.
2. Lack of Context and Domain Understanding
Generative AI is impressive at surface-level tasks: summarising, formatting, rephrasing. But it lacks context.
Ask it to write a job description? Sure.
Ask it to interpret sector-specific regulations or evaluate project risk? Expect vague generalities or hallucinated detail.
This makes generative AI poorly suited to specialised industries like construction, heritage conservation, or compliance-heavy sectors, where domain-specific reasoning matters more than tone or fluency.
Without integrating contextual frameworks, or pairing the model with human expertise, generative tools will struggle to meet the precision these sectors require.
3. Data Privacy and IP Concerns
For any organisation handling sensitive information, client records, project data, internal processes, generative AI introduces risk.
Where is your prompt data going?
Is it being stored or used for future training?
Can you use these tools without accidentally exposing IP?
Many mainstream AI platforms still lack transparent answers to these questions. For SMEs, this creates a dilemma: how to benefit from AI without compromising security, compliance, or competitive advantage.
Self-hosted models or fine-tuned private instances offer a possible solution, but they require technical investment that may be out of reach for smaller teams.
4. Inconsistent Results and Cost of Errors
Generative AI is good at “first drafts” but it’s not always consistent. You can run the same prompt twice and get radically different outputs. And when it’s wrong, it’s often confidently wrong.
For SMEs, this matters. Time spent fact-checking, correcting, or discarding poor results erodes the time savings promised by the tech in the first place.
In regulated industries or public-facing roles, the cost of error is high. AI output needs to be dependable, not just convenient.
5. Integration Challenges
Many businesses don’t need another standalone app, they need AI that plugs into existing workflows, datasets, and team processes.
Unfortunately, most off-the-shelf generative AI tools aren’t built with this in mind. Integration into CRM systems, project tools, or bespoke databases often requires custom APIs or workarounds, adding cost and complexity.
For generative AI to deliver real business value, it needs to move beyond chatbots and text boxes and support seamless, secure, and customisable deployment.
Final Thoughts:
Progress Without Illusion
Generative AI is not going away, and nor should it. But we owe it to the organisations adopting it to move beyond the marketing narrative and focus on where it truly fits, where it still needs work, and where it shouldn’t be used at all.
For SMEs and sector specialists, success with AI isn’t about being first, it’s about being smart, strategic, and selective.
