Is AI Killing the Value of 'Pretty Good' Work? How UK Professionals Can Stay Competitive

AI and the “pretty good” middle: is it being squeezed?

A sharp post on Reddit asks a blunt question: is AI quietly killing the value of being pretty good at common knowledge work – writing, research, design, coding, analysis, editing, planning?

Feels like AI may be compressing the value of that middle faster than people want to admit.

That tension is real. AI systems now produce “good enough” outputs at speed and near-zero marginal cost, which naturally devalues routine tasks. But this isn’t the end of the middle. It’s a reshuffle of where the value sits – away from generic execution and towards context, judgement, integration, and trust.

Original thread: Is AI quietly killing the value of being pretty good?

What’s being commoditised by AI right now

Models can now draft, summarise, reformat, translate, outline, and wireframe at a passable level. For many briefs, “first-draft quality” arrives instantly. Buyers notice – and they reprice.

• Writing and editing: outlines, blog posts, product blurbs, emails, SEO pages – rapid and decent, but formulaic if unedited.
• Design and content: mood boards, concept variations, thumbnails – fast iteration, less originality without strong direction.
• Coding and analysis: boilerplate, scaffolding, unit tests, data cleaning – strong accelerators, but correctness still needs review.

In other words: AI inflates supply of generic outputs. The premium shifts to specificity (your data, your users, your constraints) and accountability (someone who’ll own the result).

Where “pretty good” still holds value – with a twist

Being solid still pays when it’s paired with domain context, live collaboration, or outcomes that actually move a metric. Three anchors keep humans in the loop:

• Context and constraints: Clients rarely want text; they want the right text for a channel, tone, compliance regime, and timeline. That tailoring isn’t generic.
• Verification and responsibility: AI can hallucinate and overconfidently mislead. A human who checks sources and signs off is valuable.
• Original inputs: Interviews, proprietary data, stakeholder buy-in and messy business logic – this is where AI needs a guide.

UK-specific realities: compliance, IP, and market dynamics

For UK organisations – from SMEs to the NHS and regulated finance – compliance changes the calculus. Using public AI tools with personal or client data engages UK GDPR duties and risk assessments. The ICO’s guidance on AI and data protection is clear: know your data flows, legal basis, and risks.

• Data protection: Avoid pasting personal or confidential client data into unmanaged tools. Prefer enterprise features with data controls, logging, and regional storage where possible.
• Procurement and trust: Larger UK buyers increasingly ask for model choices, data retention policies, and auditability. Professionals who can answer win work.
• IP and originality: AI-generated assets can raise ownership and licensing questions. Document sources and maintain rights chains, especially in media and design.

Practical ways UK professionals can stay competitive

1) Move up the value chain: from doer to integrator

Value now clusters around problem framing, tool selection, prompt design, and quality assurance. Translate an objective into a workflow that combines AI with human judgement – then own the result.

2) Specialise with domain and data

Generic outputs face price pressure. Domain-specific work doesn’t. Pair your expertise with client or company data using private workflows or retrieval-augmented generation (RAG – a pattern that lets models search your documents during a query) to deliver answers they can’t get from the public internet.

3) Build small automations, not just prompts

Turn repeatable tasks into simple pipelines: spreadsheets, scripts, or low-code automations. For example, connecting a GPT to Google Sheets for reporting or content ops can cut hours per week. I’ve shared a practical walkthrough here: How to connect ChatGPT and Google Sheets with a Custom GPT.

4) Tighten the quality loop

Use checklists and lightweight evaluations. Require sources, run linting/tests, and add a human-in-the-loop for high-risk steps. Keep an error log and fix upstream prompts or data instead of patching downstream.

5) Productise and price for outcomes

Package services with clear deliverables, SLAs, and acceptance criteria. Offer tiers that reflect review depth, compliance assurance, and stakeholder management – not just word counts or hours.

6) Mind your privacy posture

Document which models you use, where data is processed, and how you handle deletion. If you’re a freelancer or micro-business, publish a short AI usage policy. This builds trust and often wins the tie-break.

Quick map: where AI pressure is high, and where humans keep the edge

Task area	Automation pressure	Human advantage
Generic copywriting	High	Original research, interviews, brand nuance, regulatory sign-off
Prototype design	Medium-High	User testing, accessibility, cross-stakeholder alignment
Boilerplate coding	High	Architecture, security, integration with legacy systems
Data analysis	Medium	Question framing, causal reasoning, decision-making under uncertainty

Signals your “pretty good” still commands a premium

• You routinely work with sensitive data, real revenue levers, or regulated outputs.
• Clients say “You understood the brief better than we did.”
• Your artefacts come with evidence, tests, or measurable outcomes.
• You get pulled into projects earlier – to shape, not just execute.

What to stop doing

• Competing on generic content mills or undifferentiated design gigs.
• Delivering AI-first drafts without human editing or source checks.
• Copy-pasting client data into unmanaged tools with unclear retention.
• Charging by the hour for tasks where AI compresses time to near-zero.

Bottom line: the middle isn’t dead, it’s moving

AI is compressing the value of undifferentiated execution. The work that holds its price in the UK market is integrated, contextual, and accountable. If you can combine “pretty good” craft with strong problem framing, data-aware workflows, and a clear privacy posture, you’re not racing AI – you’re using it to move up the curve.