My last newsletter on Claude really took off — and since then, I’ve been getting the question:
“We already have ChatGPT (or Copilot, or Gemini) — do we really also need Claude?”
It’s a fair question. Most companies I speak with already have at least one paid LLM subscription in place. Sometimes it’s rolled out broadly, sometimes just a few pilot users. But when a feature like the one that Claude recently launched becomes available and the other models don’t yet have it, I often get asked: Is one LLM enough? Should we have several? And if yes, how do we effectively and compliantly manage them all?
So let’s look at this together.
📢 Upcoming Events for Finance Leaders – Save the Dates!
AI Finance Club remains my favorite community for finance leaders exploring AI. If you’re not part of it yet, you’re missing out on some of the most practical, no-hype conversations about how finance teams can actually use AI.
I’m an expert in the Club — every month, I host the Fractional CFO Corner, a dedicated workshop for fractional CFOs. It’s a great way to see me regularly, share challenges, and learn hands-on AI strategies together.
Don’t miss the live session on how AI is reshaping the CFO role and the tools top finance leaders are already adopting by Nicolas Boucher, the leader of the community!
When: Friday, September 26, 2025 | 11:00 AM ET / 5:00 PM CET
Seats are limited—grab yours now!
The Balanced View: Do You Really Need More Than One LLM?
Let’s start with the fact that every company needs at least one corporate LLM subscription. Two things are now certain: first, there are clear efficiency gains from using LLMs — sometimes 3x, sometimes 10x — but always material. Second, your employees are using LLMs regardless of whether they have corporate access or not.
Your Default LLM is Non-Negotiable
What your default LLM should be depends on your company’s tech stack, compliance and security setup, and the most common activities your team performs. My advice is not to spend months debating which LLM to choose — just choose one and move forward.
If you struggle to define your default, here are some rules of thumb:
LLM | Best for | Notable strengths |
|---|---|---|
Copilot | Companies fully in Microsoft 365 | Seamless Excel, Outlook, Teams integration |
ChatGPT | Flexible, broad use cases | Strong reasoning, memory, connectors |
Claude | Document-heavy workflows | File handling, artifacts, structured outputs |
Gemini | Google Workspace environments | Deep Gmail, Docs, Sheets integration |
To make the most out of your subscription, you need to set it up properly:
Put security and data controls in place
Enable the right features
Set up memory and custom instructions with company context
Connect the necessary integrations
Train users continuously, not just once
Done correctly, these steps dramatically improve the quality of answers and outputs your team gets from the LLM.
When a Second LLM Makes Sense
You don’t necessarily need to give everyone access to two or LLMs, but there is value in having a second one available for AI champions or advanced users.
LLMs are changing rapidly, and features appear in one before the others. Having a second model lets you compare outputs and flag when one is stronger for a particular task.
It also prevents over-dependence on a single vendor. Some users may also need larger context windows or higher limits for research and deep analysis, so equipping them with the right tool can make a difference.
If you are considering an AI automation project — not just day-to-day LLM use — testing models in advance is critical before committing to one.
A handful of experimenters can keep your organization ahead while most employees stick to the default.
Free Tools — With Guardrails
Free tools can add value if you set clear boundaries. Some to consider are:
Google NotebookLM to organize and navigate knowledge bases, and even create podcasts or video presentations
Perplexity for external research, especially for employees who don’t have access to the corporate LLM
But remember that free does not mean safe. Your policy must define exactly what data is permitted and what must never leave your secure environment.
So, do you need more than one LLM?
For most companies, the balanced answer is:
Yes — to one default LLM. It should be properly configured, governed by policy, and supported with ongoing training so everyone knows how to use it safely and effectively.
Maybe — to a second LLM. Keep this selective, for advanced users or AI champions who can test new features, compare outputs, and flag when it makes sense to expand.
Yes — to free AI tools. But only with guardrails. Define what they can be used for and what data must never leave your secure environment.
Formalize your LLM setup, define a policy, train your users, and keep AI on your leadership agenda. That’s how you capture the value today and stay ready for tomorrow’s changes.
Deeper Dive: The Memory Setup Checklist
Memory is one of the most underutilized LLM features. Done right, it makes the model act like a real assistant for your company. Done wrong, it either produces generic outputs or hallucinates more.
Here’s how to set it up the right way (with examples for a mid-size professional services firm).
What to Include in Memory
Company-specific context
Company size, industry, and client profile
Example: “We are a mid-size professional services firm with 200 employees, serving clients in healthcare and legal sectors.”Reporting cadence
Example: “We prepare monthly financials by the 10th, quarterly board packs, and annual audits.”Chart of accounts structure
Example: “Our GL has five divisions: Consulting, Advisory, Training, Technology, and Overhead.”Billing and client engagement model
Example: “Projects typically run 6–12 months with milestone billing; average invoice size is $50–100k.”
Role-specific context
Primary responsibilities
Example: “I am Head of FP&A. I manage the annual budget, build monthly forecasts, and run scenario planning for client revenue. I prepare materials for quarterly board meetings.”
Example: “I am a Fractional Controller. I close the books monthly for four clients, each in different industries — SaaS, legal, healthcare, and consulting. I handle reconciliations, compliance checks, and prepare audit schedules.”Preferred formats and tone
Example: “For board reports, summarize in 3–4 key points with financial highlights. For staff training, use step-by-step guidance in plain language.”Work style preferences
Example: “When drafting memos, I prefer bullet points over paragraphs. When analyzing variances, highlight top 3 drivers first, then secondary details.”
What Not to Include
Personally identifiable information (SSNs, employee records, client contacts)
Detailed payroll or banking data
Full legal agreements
Regulated client data (patient records, confidential client files)
Setup Hack
If you’re unsure how to frame custom instructions, let the LLM help you. Upload a recent board deck or your company website and ask it to draft company-specific instructions. Then, in a separate step, describe your role in your own words — for example, “I am Head of FP&A, responsible for forecasting and board reporting” — and have the LLM generate role-specific instructions based on that.
Keep It Updated
Memory isn’t “set it and forget it.” Review and refresh it at least quarterly. Roles evolve, clients change, and financial cycles shift. Outdated memory leads to outdated outputs.
Closing Thoughts
The biggest risk isn’t choosing the “wrong” LLM. It’s letting employees experiment without guidance. One corporate LLM, clear memory setup, and regular training will take you further than endless tool-chasing. Add a second model selectively, and you’ll stay ahead without overspending.
The AI world is changing so quickly that having a second LLM can help you stay on top of new features and see what’s possible. But it also means carving out time and responsibility for someone on your team to track those changes — otherwise, the opportunity will be missed.
We Want Your Feedback!
This newsletter is for you, and we want to make it as valuable as possible. Reply to this email anytime with your questions or topics you’d like covered — your input shapes our content!
Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation!
Did you find this newsletter helpful? Forward it to a colleague who might benefit!
Until next Tuesday, keep balancing!
Anna Tiomina
AI-Powered CFO