The AI stack behind ODB Growth (and what it actually does)

Every second article on LinkedIn tells you which AI tools to use. Most of them list twenty options and expect you to figure out the rest.

So here is a different approach. This is the exact AI stack founder-led sales consultancies like ODB Growth run on in 2026. Nine tools. Each with a specific job. Each earning its place. Because AI reflects the discipline of the person using it, and the wrong stack amplifies the wrong discipline.

The companies winning right now don’t have more AI. Instead, they have AI that fits their process.

Why most AI stacks fail

Most sales teams collect AI tools the same way they collect CRM fields. For example, they add one because a competitor uses it. They add another because a webinar sold them on it. Furthermore, they add a third because someone on the team wanted to try it.

Six months later, they own subscriptions to eight tools. However, only two get used consistently. So the rest are decoration, and they pay for all of them.

That’s not an AI stack. Instead, that’s a graveyard of good intentions.

Most teams also don’t know how deep the gap really is. They assume they’re at a 60 out of 100 on AI integration, because everyone uses ChatGPT. In practice, when integration is measured across every function independently, the score sits closer to 34. Using tools and integrating AI into the workflow are two very different things.

A real AI stack does three things. It has a specific job per tool. It integrates with how you already work. And it pays for itself in time saved or revenue created.

“You don’t need every AI tool. You need the ones that fit how you actually work.”

The AI stack behind ODB Growth

Here is what runs my week, in the order I use it. Nine tools. Each with one clear job. Together they let one person deliver what most agencies take a team to deliver.

1. Claude (Anthropic): the strategic thinking partner

For every meaningful piece of writing, strategy or client work, I start with Claude. It thinks in structured arguments, holds long context, and pushes back when a plan is weak. Furthermore, it respects tone consistently once briefed.

Job: strategic thinking, long-form writing, client analysis, frameworks. Where a normal writer stares at a blank page for an hour, Claude produces a working draft in ten minutes.

2. Claude Code: building what I sell

For technical work, from custom CRM workflows to landing page builds and internal tools, I use Claude Code. It reads the codebase before it writes. It handles multi-file changes without breaking context. So I ship in an afternoon what previously required a developer for a week.

Job: hands-on technical execution. Every client integration, every Make.com scenario, every custom CRM automation runs through here. Building what I sell, not just talking about it.

3. Claude Cowork: for parallel task streams

When multiple client threads run at the same time, Claude Cowork handles the parallel workload. Each client, each project, each research thread gets its own context. Therefore no cross-contamination between engagements.

Job: multi-client context management. One founder handles what previously required a project manager plus three specialists.

4. Claude Design: visual work at production speed

Landing pages, one-pagers, presentation slides, brand assets. All the visual output that traditionally means either paying a designer or producing something amateur. Instead, Claude Design produces production-quality work at draft speed.

Job: brand-consistent visual assets. Client deliverables that look like they came from a team when they came from one person and one tool.

5. Claude Routines: automating the repeatable

For everything that repeats, weekly reports, content scheduling, newsletter drafts, monthly reviews, Claude Routines handles the recurring workload. It runs on schedule, produces the output, and lets me check the work instead of doing it from scratch.

Job: recurring workflow automation. Consultancy scales badly because the founder does everything. Routines removes that ceiling for predictable work.

6. ChatGPT (OpenAI): the second opinion

For every high-stakes piece of thinking, I run it past a second model. Different models catch different things. Furthermore, ChatGPT often sees the commercial angle that Claude underweights. So the disagreement between them sharpens my thinking faster than either one alone.

Job: second opinion, alternative framing. Two models beat one for anything that matters.

7. Gemini (Google): deep research at scale

When I need real research, market data, competitor analysis, or regulatory context, Gemini’s deep research mode does in twenty minutes what used to take a research analyst a day. Furthermore, it synthesises sources rather than just listing them.

Job: research and synthesis. Every client engagement starts with industry context, and this is how I get there fast.

8. Grok (xAI): real-time signal from X

For anything time-sensitive, breaking news in a client’s industry, public sentiment shifts, or emerging conversations, Grok pulls signal directly from X. So I know what’s being said today, not what was true six months ago when a training run ended.

Job: real-time market intelligence. The one tool for content and context that is not stale.

9. Cursor and Obsidian: the working environment

Not exactly AI tools, but the environment AI runs in. Cursor is where technical work happens with Claude embedded in the editor. Meanwhile, Obsidian is my second brain: every client note, every framework, every recurring pattern lives there and stays searchable across years.

Job: the operating system. Every AI tool above hooks into these two. Without them, the stack collapses.

What this stack actually delivers

This is not a list of tools. Instead, it’s a system. Together, these nine tools handle:

For example, client research on a new industry vertical in twenty minutes. Furthermore, custom CRM automations built in an afternoon. Plus landing page builds in one day. Meanwhile, content production running on a weekly cadence without me touching it. And every strategic question tested through two models before it becomes a client recommendation.

Therefore, one founder delivers what agencies bill three people to produce. Not because I’m faster than three people. Instead, because this stack removes the friction that slows three people down.

The rule that keeps this working

Every tool in this stack has one job. If a tool tries to do everything, it does nothing well. Furthermore, if a tool doesn’t earn its place every month, it goes. Plus, if two tools do the same job, one goes.

That’s the discipline that separates a working AI stack from an expensive tool collection.

Where does your AI integration actually stand?

Reading someone else’s AI stack is useful. Measuring your own is more useful.

Before you copy any of the tools above, know your starting point. This is where the AI Integration Scan comes in. It scores you across the exact bucket range and dimensions we designed the scan around.

Are you DORMANT (0-20), where AI is essentially absent and the business still runs on human hours for work AI could already do? WAKING (21-40), where a few people use AI on their own but nothing is wired in and the value leaves when they do? WIRED (41-60), where AI assists real work across several functions but still sits beside your manual processes instead of running them? AUTOMATED (61-80), where AI is embedded in the core workflows of most functions and connected to the live data underneath? Or AUTONOMOUS (81-100), where AI runs end-to-end work within guardrails and starts to shape what you sell?

Most teams can’t answer that question honestly. Furthermore, most don’t know that the seven dimensions where AI integration stalls are broadly the same across companies. Without measuring, you can’t tell whether adding another tool solves your gap or widens it.

That’s why we built the AI Integration Scan. Thirty-five questions, eight minutes, one score across the seven dimensions where AI adoption typically stalls. You’ll know exactly where your integration stands. Then you can decide whether to copy this stack, cut half of your current one, or build something entirely different.

Take the scan at odbgrowth.nl/ai-scan. No consultant required to know your number.

Practical checklist: is your AI stack actually working?

1. Every AI tool you pay for has a specific job you can name in one sentence ☐
2. You can name the last time you used each tool without checking your subscriptions ☐
3. No two tools overlap significantly in what they do ☐
4. Each tool integrates with your existing workflow, not sits beside it ☐
5. You review the stack quarterly and cut what does not earn its place ☐
6. The stack works for how you actually operate, not how influencers say you should ☐
7. New tool additions require you to cut an existing tool, not just add ☐
8. Total AI spend per month is proportional to time saved or revenue created ☐

If three or more are unchecked, your AI stack is decoration, not infrastructure. So time to audit before your next subscription renewal.

The ODB Way

At ODB Growth, we don’t recommend an AI stack. Instead, we design one that fits your process.

So we start by mapping how your team actually works. Where time goes. Which decisions get delayed. What information the team hunts for that is already available. Then we identify the AI stack that removes friction from those specific patterns.

Not a generic recommendation. Not a copy of someone else’s stack. Instead, the specific set of tools that fits how your team already operates.

Because the difference between an AI stack that works and one that doesn’t is not the tools. Instead, it’s the fit. And fit only comes from looking at your process first, not the market’s tool list.

Plus, an AI stack that doesn’t fit your process becomes one of the four sales leaks every B2B company has: another thing on the to-do list that never actually gets used. Therefore, the goal isn’t more tools. It’s the right tools, working together, delivering measurable output every week.

The first move is not adding a tool. Instead, it’s measuring where you actually stand. Then designing the stack that fits your integration level. Start with the AI Integration Scan. Everything else follows from there.