Guides
Checklists for buying, comparing and deploying, with sources.
What to confirm before buying WorkBuddy
Which Enterprise edition, what to prepare, how far the credits go and what the subscription leaves out.
InsideChoosing among the three Enterprise editions · Six things to prepare for a quote · How far the credits go · From inquiry to go-live · Four things buyers miss
Read guideEnterprise AI applicationsA one-page enterprise AI brief
The task, the data, acceptance criteria and the owner.
InsideWhat belongs on the page · A filled-in example · Choosing the first task · Designing the pilot · Four ways briefs go wrong
Read guideEnterprise AI applicationsData and compliance checklist for enterprise generative AI
What to check before adopting generative AI in China — which rules apply, personal information, content labels and supplier terms.
InsideWhich situation are you in? · Personal information: the articles to check · Content labeling · What to confirm with suppliers · Internal usage rules
Read guideEnterprise AI applicationsWorkBuddy renewals, seat changes and expiry
What to do around WorkBuddy Enterprise expiry — the pre-renewal review, what expiry stops, seat changes and invoices.
InsideSix checks a month before expiry · What expiry stops · Seat changes and reclaiming seats · Invoices · Three things teams miss
Read guideModel services and supplyModel API pricing beyond the headline rate
Billing method, input/output mix, limits and settlement.
InsideSix comparison dimensions · Filling in the comparison sheet · Cost per acceptable result: a worked example · Running a small comparison · Four common pricing mistakes
Read guideModel services and supplyWhat to bring to a model supply discussion
Resource scope, supply rights, interfaces and settlement.
InsideSix items for a first conversation · A sample resource description · Validation and next steps · Reconciliation · Four common problems in supply deals
Read guideModel services and supplyMoving to another OpenAI-compatible API
What to change and what to test — base URL, key, model name, streaming, tool calls and errors.
InsideThe three changes in code · Feature checklist · Errors and retries · Cutting over · Four common migration slips
Read guideModel services and supplyEstimating tokens and monthly cost
What a request's tokens are made of, how to measure them, how to project a monthly cost and how to lower it.
InsideWhat a request is made of · Measure, don’t guess · Projecting monthly cost · Common ways to lower cost · Four common estimating mistakes
Read guideInfrastructure capabilityHow to describe infrastructure needs
Workload type, model size, duration and data location.
InsideSix groups of requirements · A sample workload request · Choosing a deployment form · Turning the request into a proposal · Four things requests leave out
Read guideInfrastructure capabilitySplitting data-center responsibilities
Hardware, access, backups and exit: who owns each.
InsideConfirm responsibility by layer · A responsibility matrix · Changes and acceptance · Exit checklist · Four places the split goes wrong
Read guideInfrastructure capabilitySizing GPU memory for a self-hosted model
Estimate memory for weights, KV cache and headroom, weigh quantization, then measure on real hardware.
InsideWhat uses the memory · Steps and a worked example · Weighing quantization · Measuring after the estimate · Four common sizing mistakes
Read guideInfrastructure capabilityTraining, fine-tuning and inference compared
How the three workloads differ in memory, storage, network and duration, and what to state for each.
InsideThe three workloads side by side · Estimating memory for training · Storage and network differences · What to state for each · Four common planning mistakes
Read guide