The AI Budget Conversation: How Much Should a $10M Firm Spend?

That's the question I get in 90% of initial conversations. And the answer is always more nuanced than my CFO clients want it to be. But let me try to give you a framework that works.

The right AI budget for your firm depends on three things: how much AI will improve your margins, how much risk you're willing to take, and how much optionality you want to preserve. The answer for a $10M consulting firm is different from the answer for a $10M law firm or accounting firm. But the framework is the same.

The Baseline: What You're Already Spending

Before you decide how much to spend on AI, acknowledge what you're already spending on manual processes. A $10M professional services firm with 50 people typically spends 15-25% of costs on activities that AI can partially automate:

• Document review and processing
• Research and analysis
• Client intake and qualification
• Proposal generation
• Data entry and management
• Reporting and summary

A rough estimate: 15-25% of billable staff time is spent on work that AI could do. At a $10M firm, that's $1.5-2.5M in labor cost annually.

Your AI budget should be a fraction of that number. If you can recoup 20-30% of that cost through AI efficiency, you're looking at $300-750K in savings annually. Your budget should be designed to capture some meaningful fraction of that.

The Three-Tier Budget Framework

Tier 1: Minimal (0.5-1% of revenue)

For a $10M firm: $50-100K per year.

What you get: exploration and one serious pilot project. You're testing whether AI works for your firm. You're not betting the farm. You might hire a contractor for 8-12 weeks to run a specific project (document processing, proposal generation, client intake). You validate whether it works. If it does, you move to Tier 2. If it doesn't, you learn and try something else.

Use this if: you're genuinely unsure whether AI is relevant to your business, or you're extremely risk-averse.

Tier 2: Realistic (1-3% of revenue)

For a $10M firm: $100-300K per year.

What you get: two major projects running in parallel, plus infrastructure investment. You're automating 2-3 workflows. You're building or licensing a system to manage those workflows. You might hire a part-time contractor (0.5-1 FTE) to manage ongoing integration and optimization.

Budget breakdown:

• API costs: $10-30K
• Integration/development: $40-100K (contractor or vendor)
• Training and change management: $20-40K
• Software/platform tools: $20-50K
• Contingency: $10-30K

Expected return: 10-20% efficiency improvement in the workflows you automate. For a $10M firm, that's $100-200K in labor cost recovery. Payback period: 6-12 months.

Use this if: you've validated that AI works for your firm, you have 2-3 clear target workflows, and you're ready to invest in real implementation.

Tier 3: Aggressive (3-5% of revenue)

For a $10M firm: $300-500K per year.

What you get: AI is becoming a strategic capability. You're automating 4-6 workflows. You're probably hiring a dedicated resource (0.5-1 FTE) who owns AI strategy. You're investing in data infrastructure and model evaluation. You might be experimenting with custom fine-tuning or building proprietary tools.

Expected return: 20-30% efficiency improvement across multiple workflows, plus strategic advantages you can't quantify yet (faster delivery, better quality, competitive differentiation).

Use this if: AI is already demonstrating value, your leadership is aligned on this being a strategic priority, and you have the operational capacity to manage a larger initiative.

The Risk Adjustment

These budgets assume moderate risk tolerance. Adjust based on your actual risk profile:

If you're conservative: Start at Tier 1. Run a 12-week pilot on one workflow. Only move to Tier 2 if you get clear ROI. This takes 6-12 months but reduces execution risk.

If you're growth-focused: Start at Tier 2. You're betting that AI is table-stakes and you need to move fast. You're accepting higher risk of failed projects, but the upside of moving first is worth it.

The Timing Question

One more thing: the cost of waiting. If you're a $10M firm and you wait 18 months to start an AI initiative, your competitors won't. They'll have 18 months of operational data showing how AI changes their margins. You'll be starting from scratch.

For that reason, I'd rather see a firm do Tier 1 now (minimal investment, maximum learning) than wait for perfect conditions and do Tier 2 two years from now.

The Measurement Framework

Regardless of which tier you choose, measure the same things:

• Hours saved per workflow (before vs. after)
• Quality metrics (error rate, accuracy, client satisfaction impact)
• Cost per transaction (API costs + human review time)
• Adoption rate (what % of eligible workflows are actually using AI?)
• Payback period (how long until the project pays for itself?)

These metrics tell you whether your investment is working. If payback period is >24 months, something's wrong. If it's <12 months, you should probably be investing more aggressively in the next project.

The Honest Version

The "right" budget for your firm is the one where you're investing enough to learn, but not so much that you can't recover if things don't work out. For most $10M firms, that's somewhere between $100-300K per year.

You can go lower if you're extremely conservative. You can go higher if you're convinced AI is existential. But that $100-300K range is where most successful implementations land.