AI & ML

OpenAI + Smart AI Workspace

OpenAI's models cover the jobs Claude isn't the right fit for — Whisper transcription, embeddings for semantic search, and cheap high-volume classification in otherwise Claude-first builds.

How I use OpenAI

OpenAI's models cover the jobs Claude isn't the right fit for. Whisper handles audio transcription for call and meeting workflows; the embedding models power semantic search over your documents; and GPT models are a useful second opinion when I want model diversity in a critical step.

Most builds I ship are Claude-first, but I'm not religious about it. If a workflow needs cheap high-volume classification, image generation, or speech-to-text, OpenAI is often the practical choice and I'll use it for those steps.

As with Claude, the API key stays in your account, and I keep prompts and model choices documented so you can see what's running where.

OpenAIis rarely the whole system — it's one connected piece of a workflow I scope, build, and run end to end. See how I build automation or read what AI automation actually costs in 2026. Curious what your manual version costs today? Estimate it with the ROI calculator.

What I automate with OpenAI

  • Transcribe sales calls, support calls, and meetings with Whisper for downstream summarising
  • Build semantic search and retrieval over your knowledge base with embeddings
  • Run high-volume, low-stakes classification where a smaller model keeps costs down
  • Add a second model for cross-checking on decisions that carry real risk

Example workflows I build

01

Call notes to CRM

A recorded sales call is transcribed with Whisper, Claude pulls out the next steps, objections, and budget signals, and the structured summary is written to the opportunity in your CRM automatically.

02

Semantic knowledge search

Your docs, past proposals, and tickets are embedded into a vector store. A support or sales agent workflow retrieves the most relevant passages for each query and passes them to the model that drafts the answer.

03

Bulk ticket tagging

Inbound support tickets are classified by topic, urgency, and product area using a small fast model, then routed — cheap enough to run on every ticket without thinking about it.

Not sure what to automate first?

I map one or two of your processes in a free audit and tell you honestly whether automation pays off yet.

OpenAI integration FAQ

Frequently Asked Questions

Do you use OpenAI or Claude?
Both, per step. Claude for most reasoning and drafting, OpenAI for transcription, embeddings, and high-volume classification. The workflow uses whichever is right for the job.
Can I standardise on one provider?
Yes, if you'd prefer that for procurement or compliance reasons. I'll tell you where it costs you speed or money so it's an informed choice.
Where do the API costs land?
On your own OpenAI account. You see exact usage and pay the provider directly.
Is my data used for training?
API usage is not used for training under OpenAI's business terms. I configure the account and workflows accordingly.

Ready to connect OpenAI?

I'll scope the integration in a free automation audit and build the workflows around your process.