Blog
Practical articles on putting AI to work in a business: what to automate first, what it costs, what the law requires, and what we have measured in production.

Build the Claude system once, then let each country add a thin local layer. A guide to the central and local decisions, with a staged four-country example on Zoho Desk.
→
Going live with Claude is where the work starts. This guide shows how to version, test, release and roll back changes to prompts, routing rules and the Zoho data behind them.
→
Claude can sort records, draft replies and pull data out of free text inside a Zoho Creator app. This guide covers the routes, the models, the costs and the failure handling.
→
A practical guide to calling Claude from Zoho Deluge functions. It covers one parent function, API keys, model choice, reply formats, a worked CRM example and the failures to plan for.
→
Claude can draft reorder suggestions from Zoho Inventory stock levels, sales history and supplier lead times. This guide covers the setup, the maths and the buyer approval step.
→
A practical guide to using Claude with Zoho Books at month-end: what it can prepare through the connector, and which reconciliation and reporting steps stay with your finance team.
→
When Claude connects to Zoho CRM, it inherits the permissions of the user who signed in. This guide shows how to set profiles and MCP access so the AI never sees more than that person should.
→
An AI receptionist earns its keep by routing each enquiry to the right person and CRM record first time. How to design that routing, measure it and compare UK prices.
→
A practical guide to Claude drafting customer replies in English, German, Romanian and Italian inside Zoho Desk. Set the language, keep terms and tone consistent, and have a person approve every reply.
→
Before a Claude workflow touches real customers or tickets, run it against your own historical cases, edge cases and failure paths. This guide shows how to build the test set and judge the result.
→
A business case for AI that a director can approve starts with one process you already measure. This guide shows how to put costs, savings and risks on one page.
→
A builder's guide for small firms deploying an AI assistant: keep a person on significant decisions, log every AI decision, tell people, and document it all before launch.
→
Claude reads your CRM as it is. This guide shows how to find the duplicates, empty fields and stale records that would mislead it, and how to fix them before any AI project starts.
→
AI is the wrong answer when a task has one right answer, the data is thin or the process keeps changing. Here is how to tell before you spend money.
→
The real question is who wires AI into the CRM or service desk you already run. Compare building, buying and partnering on cost, ownership and exit.
→
Staff learn an AI assistant fastest on their own work, in their own tools. This guide covers AI literacy, how to ask, when to escalate and how to correct the assistant.
→
Before an AI assistant touches customer or staff data, settle roles, lawful basis, the DPIA, decision rules, access, data location and retention. This guide gives you the checklist.
→
Enterprise adds audit logs, SCIM, retention controls and compliance APIs to Team, needs at least 20 seats, and bills usage at API rates. Here is how to decide.
→
Part 1 of a series on AI agents in practice. Build a weekday Claude agent that drafts supplier replies from your CRM, keep it drafting only, and know when it needs more.
→
A practical guide to building an AI agent with the Claude Agent SDK: install it, connect your CRM through MCP, lock down permissions and keep a person in charge of risky actions.
→
An AI-native sales team writes its best methods down as skills everyone can run, on CRM data Claude can read. Here is how a company of 10 to 50 people can start.
→
Fable 5.1 costs $10 and $50 per million tokens, always thinks before it answers and stores API data in the US. What those limits change when it runs inside Zoho Desk and Teams.
→
Claude for Small Business now has 43 workflows and works with Zoho CRM, Zoho Books and Zoho Desk. This guide covers the install, the workflows worth running first and a first-week plan.
→
Claude Haiku 4.5 handles ticket triage, field extraction and CRM summaries at $1 per million input tokens. Here is where it fits, what it costs and when to step up.
→
The Microsoft 365 connector lets Claude read your mail, files and Teams chats. This guide covers the admin settings, the limits and how to wire Teams to your help desk.
→
A practical guide to splitting a business system across Claude models: what Opus 5.5 does best, what it costs, and which steps run faster and cheaper on a smaller model.
→
Claude Sonnet 5 suits most routine business work at $2 per million input tokens. This guide covers when to use it, what changes if you migrate, and where your data goes.
→
A knowledge base Claude can answer from depends on your documents more than the AI. This guide covers what to include, how to structure it, who owns it and how to test it.
→
A practical guide to email triage and drafting with Claude: the sorting rules to write, how drafts match your tone, what to escalate, what a person signs and what to log.
→
Prompt injection hides instructions in emails, files and web pages. Limit what the agent can do, block destructive actions and keep a person on every send, payment and deletion.
→
Chat volume and satisfaction scores show that an AI assistant is used, not that it works. Measure first-time-right outcomes instead, and set the baseline before go-live.
→
Opus 5.5 costs less per token than Opus 5, but four API changes can break a live integration. This guide covers the prices, the fixes and a checklist to run before you switch.
→
A decision guide to the current Claude models. Match Haiku, Sonnet, Opus and Fable to the job each step of your workflow does, then test on twenty real examples before you commit.
→
A practical guide to processing invoices and contracts with Claude: schema-based extraction for invoices, cited review for contracts, real costs, data location and the human checks that matter.
→
An AI agent is ready for production once it has three controls: human approval for consequential actions, logs that trace each decision, and a cost ceiling agreed before launch.
→
A practical guide to scoping a first AI project: one process, one measured baseline, a pilot with an agreed threshold, a running cost you can estimate and a named owner.
→
Claude can draft job adverts, summarise applications and prepare interview questions. Under UK GDPR and the Equality Act, a person must still make every hiring decision.
→
A small team needs a handful of Claude plugins, not hundreds. Start with those that connect to systems you already use, limit their rights from day one, and package your own processes privately.
→
Part 2 of our Zoho CRM reporting series covers which numbers belong in a daily briefing, how Claude should word them, how the run is scheduled and why every figure names its report.
→
Part 3 of our Zoho reporting series: how Claude drafts the weekly and monthly board report from figures your systems pull, and what stays with a person.
→
Claude should explain figures Zoho CRM has already calculated, not add up raw records itself. How to choose six or seven measures, define them once and prove each one.
→
A plain guide to building an MCP server for a home-grown app or database: when it pays off, what to expose and never expose, authentication, and the first tool to write.
→
Skip the extra storefront chatbot. This guide shows how to link Shopify or WooCommerce order data into a Zoho back office so Claude handles customer questions under your rules.
→
The Slack chat window is the easy part. This guide shows how to wire Claude into your ticketing system so internal requests are routed, logged and escalated, not just answered.
→
A plain guide to the Anthropic Academy tracks, their length and cost, a one-week plan for a small team, and the three gaps teams hit when they build a real system.
→
Power users give Claude context, reusable skills, connected systems, long tasks and a way to check its work. This guide sets out each habit and a four week path to a working system.
→
Sorting email with AI is easy to switch on. It pays only when every message reaches the right queue first time. How to compare the options, cost them and measure routing accuracy.
→
A practical guide to connecting Claude to ERP data, SAP Business One included: what to expose first, the three routes, a purchase order example and where integrations break.
→
Claude can draft the wording of a quote from your CRM deal. This guide shows how to keep every price, discount and approval rule in Zoho, so no figure the customer sees comes from the model.
→
One Project per process, Skills written as procedures, shared knowledge files with named owners, and a simple test routine. A practical setup guide for small teams using Claude.
→
A practical guide to linking Claude with HubSpot or Salesforce via MCP. It covers the connectors on offer, permissions, field mapping, and which decisions should stay with your people.
→
How to draw the line on what staff may paste into Claude. Approved plans, a data table, approvals, logging, who owns the policy, and a rollout you can finish in a week.
→
A practical guide to connecting Claude to your accounting package: what Xero, QuickBooks and Sage expose, what to grant, and what a person must still approve.
→
A practical guide to connecting Claude to Zoho CRM with MCP. It covers the four Zoho CRM servers, the connector routes, permission scoping and the checks to run before anyone gets write access.
→
A practical guide to Claude Cowork for business: plans, installation, the first three tasks to hand it, which folders and accounts it may touch, and when your CRM needs a proper integration.
→
A practical guide to a website AI assistant on Claude: what it answers from, what it must never promise, when it hands over, and how the lead reaches Zoho CRM and Bookings.
→
What Claude costs a ten person company per year depends on three lines: seats, metered API usage and setup work. This guide explains each and builds three yearly budgets.
→
Company names, rankings and day rates say little about delivery. Here is how to judge a UK AI consultancy by the one thing that counts: a working system with measured results.
→
A strategy deck proves little. This guide shows how to judge an AI consultancy by working systems, with published UK pricing, due diligence checks and a service desk example.
→
Standalone chat tools only know what you paste in. AI built into your help desk or CRM reads live records, writes results back and leaves an audit trail.
→
Claude earns its keep in a small business when it works inside the tools your team already uses. Here is what the new package covers, where it stops, and how to start.
→
Whether Claude is GDPR compliant depends less on the model than on your plan, your contract and your system design. This checklist covers what European businesses should verify first.
→Claude for business works best inside the systems your staff already use. This guide compares the packaged options, costs and data rules, then looks at a live Teams service desk that routes into Zoho Desk.
→