# Six months running my consulting practice on Claude Source: https://getstartedwithai.net/blog/six-months-with-claude Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # Six months running my consulting practice on Claude Claude vs ChatGPT for business: Claude Pro wins on long-document work, voice fidelity on drafts, and Artifact-based deliverables. ChatGPT Plus wins on speed for short tasks, image generation, and integration breadth. After six months running my UK consulting practice on Claude as the primary assistant, the honest answer is that both tools have a job and the right setup runs both. This post is the workflows where each one wins, the workflows where the choice does not matter, the £32 a month I pay for both, and the UK GDPR view that almost no comparison covers. ## What got noticeably better on Claude - Voice fidelity on long drafts. Claude produces drafts that pass for my own writing more often than ChatGPT did, even with calibration. - Long-document work. The 200K context window means a 100-page client brief stays usable end to end without summarisation tricks. - Artifacts as deliverables. Producing a downloadable .docx, .csv or HTML inside the chat changed how I run client output. Two clicks from prompt to deliverable. - Audit notes. Claude is better at producing a one-page recommendation from a two-hour transcript without flattening the practitioner voice. - Proposal drafting. Three hours of prompt, transcript and prior proposal becomes a usable draft in 15 minutes. ## What stayed the same - Short, varied work. Both tools handle a 20-message email triage session equally well. - Code review tasks. ChatGPT is still slightly stronger on the kinds of code review I do, mostly Zapier and Make logic. - Image and chart generation. Neither has caught up to Canva or Midjourney for visual work, and ChatGPT is still the faster path when I need a quick chart. - Custom GPTs. The OpenAI Custom GPT format is still the cleanest way to ship a templated workflow to a client. ## What got worse - Latency on the most-used model. Sonnet 4.6 is slower than GPT-5.3 Instant for short tasks. - Cost-per-message on heavy days. I burn through Pro limits about once a fortnight; the Max tier is the next step. - Plugin and integration breadth. ChatGPT still has a wider ecosystem of integrations, and several of the integrations I use weekly are OpenAI-native. - Real-time browsing. Claude has improved here but Perplexity Pro still wins for citations and ChatGPT still wins for fast lookups. ## Where Claude beats ChatGPT in consulting work Three workflows in my practice run on Claude every week and would not be the same on ChatGPT. The first is proposal drafting, where Claude holds the practitioner voice across a 1,200 word document without slipping into the consultancy register. The second is audit transcripts, where the long context window means I can paste a full two-hour discovery transcript plus the prior client brief plus my notes file and get a one-page recommendation in a single pass. The third is client-safe outputs, where Claude is the assistant I trust to produce a draft I can send without rewriting, on the kinds of email and letter that an SMB client expects in their own voice. ## Where Claude loses, and the workaround Claude loses on image generation, real-time browsing and the breadth of plugins. The workaround is honest: keep ChatGPT Plus alongside Claude Pro at £16 a month each, use ChatGPT for the work it wins, and stop pretending one tool will do everything. Total cost £32 a month. The saving on a single client task per week clears that within five days. Anyone selling a single-tool stack in 2026 is selling a brand affiliation, not a working setup. ## Side-by-side: when to pick each Pick Claude for any task longer than three turns, any document over 10 pages, any draft where voice fidelity matters more than speed, and any deliverable that benefits from being downloadable inside the chat. Pick ChatGPT for fast lookups, code snippets, image generation, scheduled tasks and Custom GPTs you want to ship to a client. Both tools handle the middle ground. The wasted spend is paying for two competing assistants in the same lane (Claude Max plus ChatGPT Pro plus Gemini Advanced) when the work fits inside Pro on each. ## Subscription vs API for a small consultancy For a one to three person consultancy, the subscription tier is the right starting point on both tools. Claude Pro at £16 a month and ChatGPT Plus at £16 a month covers daily client work for a single practitioner with room to spare. The API becomes the right answer when a client build needs the model running on a schedule (a daily summarisation job, an inbound email classifier) or when a Custom GPT does not fit inside the OpenAI quota. The honest test: if the API spend is going to clear £40 a month for a single workflow, lift the workflow into a Zap or Make scenario with the API on the back, and bill it to the client as part of the build. ## UK GDPR and DPA 2018 for client data Almost no comparison post covers this and it is the single most important question for a UK consultant. Both tools offer a setting to opt your account out of model training. Both Anthropic and OpenAI publish data processing addenda for business accounts. For client data that includes personally identifiable information, the right setup is the business or team tier on either tool, with training opt-out switched on, and a written processor relationship with the client documented in your engagement letter. For purely internal work (proposal drafting, your own meeting notes) the Pro tiers are fine. The wrong setup is processing identifiable client data on a personal account with default settings, which a UK practitioner running under DPA 2018 should not do. ## Build your first AI system live with Richard A 60-minute live build with Richard Batt. Bring one repeatable task. Leave with a mapped automation, success metrics and a deploy plan you can test the next morning. Free seat, no upsell. ### Live build session The next live session covers the WIRE Method end to end on a real audience build. [Join the next session](/webinar) ## Questions people ask before they start ### undefined undefined ### undefined undefined ### undefined undefined ### undefined undefined ### undefined undefined