Building an AI Agent Stack for a Small Consultancy: My Recommended Setup

Most consultancies have a chaos of tools. I visited a five-person consulting firm last year and counted their subscriptions: Slack, Microsoft Teams, Notion, Asana, Monday.com, Zapier, n8n, ChatGPT Plus, Claude, Gemini, Perplexity, Figma, Adobe Creative Suite, Google Workspace, Microsoft 365, Calendly, Stripe, Wave, and six others I've forgotten. Total spend: £8,400 per year. How many of these tools were actually essential? Three, four.

The problem is that consultants, like most teams, add tools opportunistically. Someone reads about a new AI assistant and gets an account. A client requires a specific collaboration platform. A process breaks and you grab the first tool that fixes it. After 18 months, you have 15 subscriptions, half of them unused, none of them integrated.

I've spent the last five years working through this problem across 120+ consulting engagements. What I've learned is that a small, well-integrated stack beats a large, fragmented one every time. This post is my personal recommendation for a minimal, coherent AI agent stack for a consulting business.

The Core Stack: Five Tools That Actually Matter

Here's what I use and recommend to consulting teams:

1. Claude (via API) as your primary AI agent and coding assistant. This is non-negotiable in my setup. Claude's reasoning is the strongest, its coding abilities are excellent, and it's easy to integrate. I use it for nearly everything: analysis, writing, coding, brainstorming, client communication drafts. Cost: £20/month for the web interface, £0.01-0.10 per request via API depending on usage.

2. Claude Code for custom automation and data processing. When I need to build something: a report generator, a data processor, a file workflow, I use Claude Code directly. It's faster than writing scripts myself and integrates smoothly with Claude's broader reasoning. I probably use this 3-4 times per week on client work. Cost: included in Claude subscription.

3. Zapier (or n8n if you self-host) for workflow automation. Not everything needs custom code. Connecting tools, automating routine sequences, triggering actions based on events. Zapier handles this faster than building custom code. I use it for: automatically logging client emails to notes, triggering reports when data arrives, syncing between tools. Cost: £20-40/month for mid-tier use.

4. Airtable as lightweight database and operations hub. For anything that needs structure but not full software engineering: client projects, deliverable tracking, resource planning, I use Airtable. It's flexible, integrates with everything via Zapier, and my team can understand it without being technical. Cost: £20/month for a team account.

5. Google Workspace for documents, email, and collaboration. I use this for everything involving text and collaboration: proposals, documents, email, calendaring. If you need Microsoft for a specific client, fine, but standardise on one ecosystem. Cost: £10-14/person/month (usually already paid for by most consultancies).

That's it. Five core tools. Total cost for a three-person consultancy: roughly £150-180/month for new subscriptions. (Most consultancies already pay for Google Workspace.)

Everything else is built on top of these or is client-specific. Not 15 half-used tools. Not subscriptions you forgot about. Five tools that do 95% of what you actually need.

What You're Not Using (And Why)

Let me be explicit about what's deliberately not in this stack, because consultants often ask why.

Why not dedicated project management software? I tried Asana, Monday.com, Linear, and others. For small consulting teams, they're overkill. Airtable with a few views and linked records handles projects, timelines, and deliverables perfectly well. The moment you add Asana, it becomes another place data lives instead of one source of truth.

Why not separate brainstorming and ideation tools? Miro, FigJam, and others are nice-to-haves. Honestly, a Google Doc or whiteboard works fine. Consultants often confuse having a tool with doing better thinking. They don't correlate after a point.

Why not ChatGPT, Gemini, Perplexity, etc.? I have accounts for all of them. But they're not core to my workflow. Claude via API and the web interface covers 98% of what I need. The others I use once a month. Don't pay for every AI assistant: pick one that works for you and go deep with it. Switching context between tools costs more than it's worth.

Why not dedicated customer relationship management software? Most consulting CRMs are built for sales organisations, not service delivery. I track clients in Airtable with linked projects and deliverables. It's lightweight and I can customise it in an hour. The moment you move to a "proper" CRM, you've added another system to maintain and another place where data gets stale.

Why not Slack instead of Google Workspace for chat? Nothing against Slack. If your clients all use Slack, use it. But integrating email, chat, and documents into one ecosystem (Google Workspace) reduces context-switching. You read an email in Gmail, draft a response in Docs, and everything is searchable in one place. Slack + email + documents = three systems to search.

The Second-Tier Tools (When You Actually Need Them)

Once the five core tools are embedded, there are a few specialist tools worth adding for specific use cases:

Perplexity or Bing for research and up-to-date information. Claude doesn't have real-time web access. When I need current information or recent research, I use Perplexity. Cost: free tier covers 90% of needs. £20/month for unlimited queries if needed.

Figma for any design or wireframing. Only if you're doing design work. Consultancies that don't do design work don't need this. But if you do, Figma is the standard. Cost: £12/month individual.

NotebookLM or similar for converting documents into knowledge bases. Occasionally useful for quickly building a knowledge base from client documents. Not essential. Cost: varies, often free tier is sufficient.

Loom for video recording and explanations. When a text explanation won't do, Loom is faster than any alternative. Use it when clarity matters more than written documentation. Cost: £10/month.

These are genuinely optional. Add them only when you find yourself needing them regularly. Don't speculate.

The Actual Workflow: How This Stack Connects

Here's how I actually use these tools in a typical week of consulting work:

Monday morning: Check Airtable for the week's projects and deliverables. This is my single source of truth for what's happening.

Client communication: Email arrives in Gmail. I draft responses in Google Docs if they're complex, use Claude to refine them, send them back. Everything stays in Gmail for history.

Analysis work: Client sends data (CSV, Excel, PDF). I use Claude Code to load it, explore it, run analysis, and generate a formatted report. I email the report. Done in 2-3 hours instead of a full day.

Routine automation: Client has a recurring process (weekly report, data sync, client update). I set it up in Zapier if it's simple (moving data between tools, sending emails on triggers). If it's complex, I use Claude Code to build a custom script.

Project tracking: As work progresses, I update Airtable. Linked records mean everything that touches a project automatically stays in sync. If a client asks for a project status, I pull it from Airtable: one source of truth.

Brainstorming and ideation: Thinking through a strategy question or design challenge? Google Doc + Claude. I write freely, Claude helps me organize and refine. No separate ideation tool needed.

Knowledge base: Building a knowledge base from client documentation? Use Claude Code to batch-process it, then store it in Airtable or Google Docs with proper linking.

This entire flow uses five tools. Everything connects. Information flows naturally. I'm not context-switching between incompatible systems.

Why This Stack Works Better Than Alternatives

I've seen consultancies with 15+ tools that are less productive than ones with 5. Here's why this specific combination works:

Single source of truth for projects: Airtable is the hub. Everything ties back to it. Clients are projects. Deliverables are projects. Resources are linked to projects. No disconnection between your project view and reality.

AI integrated, not bolted on: Claude isn't a separate tool you open in another tab. You use Claude to think and draft, and you use Claude Code for anything technical. They're part of your core workflow, not distractions.

Minimal switching: Your email, documents, and collaboration happen in Google Workspace. Your projects and structure live in Airtable. Your automation happens in Zapier or Claude Code. That's it. You're not switching between 10 different systems.

Scalable without adding complexity: A one-person consultancy and a 10-person consultancy can use this exact stack. You just adjust the Airtable views and who has access. You don't need enterprise versions of complicated tools.

Adaptable to clients: Clients might require specific tools (they want Asana, they use Salesforce, they need Teams). Fine. That's client-specific. Your internal stack stays consistent. You integrate with their tools via Zapier when needed.

The Tooling Mistakes I See Consultancies Make

Before I wrap up, let me point out the most common mistakes:

Mistake 1: Adopting tools to feel modern. Consultants get excited about a new tool, build it, then it becomes one more thing to maintain. Only adopt tools that solve actual problems you have today, not hypothetical problems you might have someday.

Mistake 2: Letting tools proliferate without integration. Once you have five tools, integrate them. If Airtable and Zapier aren't connected, you're not getting the value. Data should flow between tools automatically, not manually.

Mistake 3: Tool complexity creeping up. Someone learns Airtable deeply and starts building complex automations in it. Suddenly it's unmaintainable. Sometimes the simple answer (Claude Code script) beats the "elegant" answer (complex Airtable logic).

Mistake 4: Ignoring the cost of context-switching. You think Slack is cheap and Perplexity is cheap and Loom is cheap. But if you're switching between 10 tools per day, the context-switching cost (in focus and productivity) exceeds the tool costs.

Mistake 5: Not standardising within the team. If you have two team members using different stacks, nothing integrates. Data lives in multiple places. Onboarding new people becomes harder. Standardise and everyone is more productive.

Getting Started: A Practical Checklist

If you're building a consultancy stack from scratch:

Week 1: Set up Google Workspace for email and docs (you probably already have this). Get Claude (web access or API depending on your needs). Cost: £20 if you're adding just Claude.

Week 2: Set up Airtable with a basic structure: clients, projects, deliverables, team members. Spend a day designing it properly. Cost: £20/month.

Week 3: Connect Airtable to Google Workspace via Zapier (automatically create docs for projects, log emails into a client record). Cost: £25/month for Zapier.

Week 4+: Use Claude Code for custom automations as needed. Refine your Airtable structure based on what you actually need (not what you think you might need).

Total onboarding time: 20-30 hours. Total monthly cost: £60-75. Total impact: 10-15 hours per week freed up from administrative work.

One Year In: Evolution, Not Revolution

Eighteen months ago, I was using a different stack. I had Notion, some custom scripts, and a CRM. Then I got frustrated with the fragmentation and rebuilt around these five tools. It took a month to migrate, another month to stabilise, and then I've been running smoothly.

What I've learned is that the best stack is the one you'll actually use and maintain. Complexity is the enemy. Fancy features you don't need are the enemy. Your stack should enable your thinking and your delivery, not become another thing to manage.

I've used this stack with consulting teams of 1-15 people. It scales. It works. It integrates. And it costs about a tenth of what most consultancies are spending on fragmented tooling.

If your consultancy is drowning in tools or you're just getting started and want to avoid that trap, let's design a coherent stack that fits your specific needs and integrates properly. I've optimised this across 120+ consulting engagements. Let's talk here about building a toolkit that actually supports your practice.

Richard Batt has delivered 120+ AI and automation projects across 15+ industries. He helps businesses deploy AI that actually works, with battle-tested tools, templates, and implementation roadmaps. Featured in InfoWorld and WSJ.

Frequently Asked Questions

How long does it take to build AI automation in a small business?

Most single-process automations take 1-5 days to build and start delivering ROI within 30-90 days. Complex multi-system integrations take 2-8 weeks. The key is starting with one well-defined process, proving the value, then expanding.

Do I need technical skills to automate business processes?

Not for most automations. Tools like Zapier, Make.com, and N8N use visual builders that require no coding. About 80% of small business automation can be done without a developer. For the remaining 20%, you need someone comfortable with APIs and basic scripting.

Where should a business start with AI implementation?

Start with a process audit. Identify tasks that are high-volume, rule-based, and time-consuming. The best first automation is one that saves measurable time within 30 days. Across 120+ projects, the highest-ROI starting points are usually customer onboarding, invoice processing, and report generation.

How do I calculate ROI on an AI investment?

Measure the hours spent on the process before automation, multiply by fully loaded hourly cost, then subtract the tool cost. Most small business automations cost £50-500/month and save 5-20 hours per week. That typically means 300-1000% ROI in year one.

Which AI tools are best for business use in 2026?

It depends on the use case. For content and communication, Claude and ChatGPT lead. For data analysis, Gemini and GPT work well with spreadsheets. For automation, Zapier, Make.com, and N8N connect AI to your existing tools. The best tool is the one your team will actually use and maintain.

Put This Into Practice

I use versions of these approaches with my clients every week. The full templates, prompts, and implementation guides, covering the edge cases and variations you will hit in practice, are available inside the AI Ops Vault. It is your AI department for $97/month.

Want a personalised implementation plan first? Book your AI Roadmap session and I will map the fastest path from where you are now to working AI automation.