Is AI Killing the Value of Being ‘Pretty Good’ at Knowledge Work? How to Stay Valuable

AI and the shrinking value of being “pretty good” at knowledge work

A thoughtful Reddit post asks a blunt question: is AI quietly killing the value of being pretty good at things like writing, research, design, coding and analysis? It’s a fair concern – the middle is where most of us live, and where AI is now strongest.

“The huge middle ground where being solid enough used to have real market value… Feels like AI may be compressing the value of that middle.”

Here’s what that means in practice for UK developers, freelancers and knowledge workers – and how to stay valuable as the bar shifts.

Source: Reddit discussion

What AI does to “pretty good” work

Modern AI systems (especially large language models trained on vast text corpora) are built to generalise. They produce reasonable first drafts, working code snippets and competent analysis very quickly. That’s exactly what “pretty good” used to mean.

• Baselines are now near-instant. Drafting, summarising, boilerplate code, slide outlines and data clean-up can be generated in seconds.
• Speed and consistency beat average output. If a client needs a decent report or landing page by 4pm, AI often delivers “good enough” faster and cheaper.
• Quality is compressing towards the middle. The gap between an average human and an AI baseline is narrowing across many text- and pattern-based tasks.

That doesn’t make skilled people redundant. It does change where value accumulates. Being the person who types the first draft or writes routine functions is less valuable. Designing the brief, evaluating the output, integrating with business context and owning the outcome is more valuable.

Where “pretty good” still matters – and where it doesn’t

Commoditised tasks: high risk of displacement

• Generic copy, emails, basic research, transcription, simple CRUD code, standard image edits, spreadsheet formulas.
• Work that can be judged syntactically (does it compile? is it grammatical? is the image centred?).
• Tasks with low domain risk and weak feedback loops.

Durable edges: lower risk, higher value

• Ambiguous briefs, trade-offs, and accountable decisions where context matters (budget, regulation, brand risk, stakeholder politics).
• Integration work: stitching tools, data and processes together reliably, with security and audit in mind.
• Regulated or high-stakes domains (health, finance, legal, safety) where accuracy, provenance and sign-off are non-negotiable.
• Trust-based client work: workshops, discovery, negotiation, facilitation, and post-mortems.

Practical moves to stay valuable in the UK market

1) Become an AI power user, not a passive user

• Build repeatable workflows: prompt libraries, style guides, checklists and test sets for your niche.
• Automate the boring bits: connect models to your data and tools (sheets, docs, Slack, Git). If you live in spreadsheets, see my guide on connecting ChatGPT to Google Sheets.
• Ship with guardrails: enforce review steps, keep versioned prompts, and document known failure modes.

2) Move up the problem stack

• Own the brief and outcomes: define success metrics, constraints and acceptance tests before generating anything.
• Offer design, not just delivery: discovery, user research, roadmaps, experiment design, and cost-risk analysis.

3) Specialise vertically

• Pick a domain where context is king (e.g. UK financial services, NHS data, construction, energy, education, charity sector).
• Learn the codes, standards and data shapes those clients use. Your edge is knowing what “good” means in their world.

4) Bring your own data and processes

If you can integrate a model with proprietary or client-specific sources, your output stops being generic.

• Curate internal knowledge bases, glossaries and templates.
• Use retrieval-augmented generation (RAG) – a method that lets models cite and ground answers in your documents – to improve reliability and auditability.
• Keep governance in mind: document sources, update cadence and permissions.

5) Price outcomes, not hours

• AI shrinks task time. If you price by the hour, you punish your own efficiency.
• Switch to retainers, milestones and performance-based fees where possible.

6) Sell trust: compliance, privacy and provenance

• UK GDPR applies. Be explicit about data handling, retention, model providers and where data is processed.
• Offer AI usage policies, DPIA support and audit trails. See the ICO’s guidance on AI and data protection.

A quick map: where AI helps and where humans win

Task	AI Baseline	Human Edge
Drafting reports/articles	Fast, coherent first drafts	Angle, audience fit, sourcing and sign-off
Coding boilerplate	Scaffolds, tests, refactors	Architecture, security, performance trade-offs
Data analysis	Summaries, charts, SQL	Correct framing, causality, error checks, decisions
Design	Variations, assets, layouts	Brand systems, UX research, accessibility
Project planning	Templates, timelines, risks	Stakeholder management, dependencies, trade-offs

UK-specific considerations: risk, privacy and procurement

• Data protection: If you handle personal data, document lawful bases, minimisation and retention. Some providers offer enterprise controls and data residency – scrutinise these before use.
• Confidentiality: Many UK clients (especially in finance, health, public sector) ban external AI tools without a data processing agreement. Offer on-prem or private deployments where needed.
• IR35 and contracting: If AI lets you deliver more value with fewer hours, position yourself around outcomes and ongoing risk management, not time-and-materials.
• Accessibility and standards: Public sector work must meet WCAG and procurement rules. Bake these into your AI workflows and QA.

Signals your role is vulnerable vs resilient

Vulnerable

• Your output is generic and judged only on speed and surface quality.
• Little exposure to stakeholders, strategy or live data.
• No measurable link between your work and business outcomes.

Resilient

• You define problems, not just implement solutions.
• You control data, integration, evaluation and sign-off.
• You’re accountable for results and can explain decisions to non-technical stakeholders.

Bottom line: the middle is moving, not disappearing

AI is compressing the value of “pretty good” execution. That middle now belongs to people who pair AI with domain context, control data and processes, and take responsibility for outcomes. If you’re the person who can turn a messy brief into a reliable, auditable, compliant solution – with AI as leverage – your value goes up, not down.

The opportunity is to stop competing with AI at the task level and start orchestrating it at the system level. That’s where the new middle sits – and it still pays.