Practical guides for AI that ships
Twenty long-form guides on how LLM systems actually work, what they cost, how they fail and how to lead them. Opinionated, vendor-neutral, written by practitioners.
- 20 guides
- 9 minutes each
- Updated November 2026
Two reading paths
The same system looks different from the boardroom and from the terminal. Pick the path that matches the decision in front of you.
Read the AI readiness guide, then build vs buy, then the AI CTO playbook. Finish with measuring AI ROI so the first project has a number attached before it starts.
Read RAG explained, then the vector database comparison, then evaluating LLM outputs. Before anything reaches users, work through the LLM security checklist.
From first principles to a measured programme
How the building blocks work
The mechanisms behind the acronyms. Read these before you sit in a vendor demo.
RAG explained
How retrieval-augmented generation grounds answers in your documents, where it breaks, and the five design decisions that matter.
9 min read FoundationsMCP explained
The Model Context Protocol: how hosts, clients and servers hand tools and data to a model, and what to expose first.
9 min read FoundationsAgentic workflows 101
The plan-act-observe loop, when an agent beats a pipeline, and the guard-rails that keep autonomy from becoming a liability.
9 min read FoundationsPrompt engineering vs fine-tuning
The ladder from prompting to RAG to fine-tuning, and why most teams should climb it one rung at a time.
9 min read FoundationsSmall language models
Where a small model beats a frontier one on cost, latency and privacy, and where the trade-off bites.
9 min readDecisions you make while building
Comparisons and checklists for the choices that are expensive to reverse.
Vector database comparison
pgvector, Milvus, Pinecone, Qdrant and Weaviate compared on the criteria that decide the choice, with a decision tree.
9 min read BuildChoosing embedding models
Dimensions, languages, domain fit, hosting and the migration cost nobody budgets for.
9 min read BuildEvaluating LLM outputs
The evaluation pyramid: unit checks, golden sets, model-graded rubrics and human review, and how much of each you need.
9 min read BuildThe cost of LLM applications
Where the money actually goes across tokens, retrieval, evaluation and people, and the levers that cut it.
9 min read BuildSalesforce and LLMs
Agentforce, Einstein and custom models: an architecture for putting language models safely inside your CRM.
9 min read BuildAI for events
Matchmaking, support bots, live transcription and analytics across the event lifecycle, from our own Eventify work.
9 min readBefore it reaches users or regulators
Security and compliance are design inputs, not sign-off steps.
LLM security checklist
Prompt injection, data leakage, tool abuse and supply chain: a layered checklist to work through before launch.
9 min read Secure & complyEU AI Act compliance guide
The risk pyramid, what counts as high-risk, obligations by role and a practical path to readiness.
9 min read Secure & complyOn-prem vs cloud LLMs
Data residency, control, cost and capability: a decision table for where your models should run.
9 min readRunning AI as a programme, not a project
For the people who own the budget, the roadmap and the answer to "was it worth it".
AI readiness guide
Six dimensions of readiness, how to score them honestly, and what to fix before the first pilot.
9 min read Lead & measureBuild vs buy AI
A two-axis matrix for deciding what to build, buy, or assemble, and the hidden costs on each side.
9 min read Lead & measureThe AI CTO playbook
A 90-day plan for the person accountable for AI: inventory, guard-rails, one production win, then scale.
9 min read Lead & measureData readiness for AI
The checklist that decides whether a RAG or analytics project succeeds before a model is chosen.
9 min read Lead & measureThe AI project roadmap
Discovery to production in phases, with the gates, deliverables and durations we actually use.
9 min read Lead & measureMeasuring AI ROI
A formula that survives the finance review, the metrics behind it and the mistakes that inflate it.
9 min readGo deeper
AI glossary
Plain-English definitions of the terms used in every guide, from embeddings to the EU AI Act.
Browse termsCase studies
How these ideas played out in Eventify, ZekoHR and client work.
See the workAI readiness assessment
A ten-minute interactive score across the six dimensions the readiness guide describes.
Score your organisationLLM cost calculator
Turn the cost guide into a monthly estimate for your own traffic and prompt sizes.
Estimate spendFrequently asked questions
Who writes these guides?
Teksmobile's AI practice — the consultants and engineers who scope, build and secure LLM systems for clients. Nothing here is syndicated or auto-generated; each guide reflects decisions we make on real projects.
How often are the guides updated?
We review every guide when a model, standard or regulation changes materially, and at least twice a year. Each page shows its last-updated date under the title.
Can I reuse this material internally?
Yes. Share links freely and quote with attribution. If you want a tailored version for your board or engineering team, talk to us about a briefing.