The 10-20-70 Rule for Small Business AI Spending

Per BCG's analysis of AI transformation, the pacesetters in AI investment follow the 10-20-70 rule: 10% of effort on algorithms, 20% on technology and data, 70% on people and processes. The ratio is well-known at enterprise scale and largely ignored at small business scale, where buyers typically reverse it: 90% on tools, 10% on everything else. Per a Forbes analysis citing BCG research, the average organization currently dedicates only 1.7% of revenues to AI investment — meaning the people-and-process portion is a meaningful slice of small budgets, not enterprise rounding error.

The reversal explains why most small business AI investments don't pay back. The tools work; the people-and-process work that turns tools into value never happens. This article is the rule applied to small business reality and how to actually run it.

What the rule means at SMB scale

For a small business spending $200/month on AI tools, the 10-20-70 rule implies:

$200/month on tools (the 10%).

Equivalent $400/month value on integration (the 20%). This is mostly your team's time configuring tools, integrating with existing systems, learning how to use them. A 5-person business spending 8 hours total per month on AI integration at $50/hour loaded cost is the equivalent of $400/month of investment.

Equivalent $1,400/month value on people and process (the 70%). This is the hours spent training the team, redesigning workflows, building habits, troubleshooting adoption issues, measuring results. Spread across the team, this is 28 hours/month total — about 6-7 hours per person at a 5-person business.

The cash spend is $200/month. The total investment is $2,000/month equivalent. Most small businesses see only the cash spend and skip the rest.

Why the people portion matters more for SMB

Three reasons the 70% is even more important at small business than enterprise scale:

Concentration of impact. In a 5-person business, every person matters. If one person doesn't use the AI, you've lost 20% of the gain. In a 5,000-person enterprise, individual non-adoption is noise. SMB AI succeeds when the whole team adopts; partial adoption doesn't produce ratio-of-effort returns.

No specialist roles. Enterprises have AI specialists who own the implementation. Small businesses don't. The owner or a designated team member has to drive adoption alongside their other work. The "people work" is invisible labor, not someone's job description.

Tighter ROI margins. Enterprises can absorb a failed AI investment. Small businesses can't. A $30K AI investment that doesn't deliver represents 1-3% of revenue for a $1-3M business. The same $30K is rounding error for a $300M business. The discipline that prevents waste matters more.

What the 10% looks like

The 10% in cash spend covers:

Subscription costs for AI tools. $20-$300/month for the working stack.

Per-use compute (rare at SMB scale). Some workflows have variable costs based on usage; usually small relative to subscriptions.

One-time implementation purchases. Sometimes a one-time setup fee for a tool or platform; usually under $5K.

For a small business spending $20K/year on AI ($1.7K/month), this is appropriate.

What the 20% looks like

The 20% in time-and-effort covers:

Setup and configuration. Each new tool needs setup. Integration with existing tools (calendar, CRM, email) takes time. Bookkeeping AI needs your transaction history; CRM AI needs your customer data; content AI needs your brand voice trained. Total: 5-15 hours per tool.

Process design. Decisions about how the tool fits into existing workflows. Who uses it. When. For what. The decisions take 2-4 hours per tool but are essential — tools without process design get used inconsistently and don't compound.

Documentation. A short runbook for each tool: what it's for, how to use, what to do when it fails. 1-2 hours per tool. Saves dramatically more time later when team members forget or new hires onboard.

For a 5-tool stack, 50-100 hours of setup-and-process time over the first 3 months. At loaded labor cost, this is $2,500-$5,000 of investment-by-time.

What the 70% looks like

The 70% in people-and-process is the hardest to spec because it's mostly invisible work. Practical components:

Team training and habit-building. 1-3 hours per team member per tool initially, plus ongoing 30-60 minutes/month per person reviewing usage and adjusting. For a 5-person team and a 5-tool stack, 25-75 hours of initial training plus 15-30 hours/month ongoing.

Workflow redesign. Existing processes that the AI changes need to be rebuilt around the new capability. The CRM with AI lead scoring needs a new follow-up workflow. The chatbot needs an escalation process. The content tool needs editorial standards. Each is 5-15 hours of design work.

Performance review and adjustment. Are people actually using the tools? Are the tools producing the expected outcomes? Monthly review of usage, output quality, and team feedback. 2-4 hours/month of leadership attention.

Troubleshooting and support. AI tools fail in surprising ways. Someone has to handle the failures. 2-5 hours/month of designated owner time.

Experimentation and improvement. Prompts that work better; integration patterns that compound; new use cases worth piloting. 3-5 hours/month of ongoing learning.

Total: 30-50 hours/month of distributed people-and-process attention for a 5-person business with a 5-tool stack. At loaded cost, $1,500-$2,500/month equivalent investment.

What happens when you skip the 70%

The visible failure modes:

Tools paid for but unused. Subscriptions on auto-renewal that nobody opens. The cost continues; the value is zero.

Inconsistent usage. Some team members use the tool, others don't. Outputs are inconsistent. The team falls back on old workflows.

No measurable ROI. Without process design and review, you can't tell if the AI is producing value. You assume it is because you're paying for it; the data wouldn't support the assumption.

Workflow regression. The team uses the tool incorrectly, fights with it, eventually abandons it. The previous workflow returns. Investment becomes net negative because it consumed attention without producing value.

Hidden quality problems. AI produces output that's "fine" but lower quality than what the team would have produced manually. Customers notice slowly. Brand damage accumulates.

These failure modes are predictable. Almost all of them trace back to insufficient 70% investment.

How to actually deploy the 70%

Practical prescription for a small business deploying AI:

Pick one person as AI owner. Not full-time; 4-8 hours/week of attention. This person drives adoption, troubleshoots, improves processes. The role can rotate every 6 months but must always be filled.

Schedule weekly 30-minute team AI review for first 8 weeks of any new tool. Discussion: what's working, what's not, who's using it, where adoption is stuck.

Document each tool's role explicitly. "We use [tool] for [specific task] in [specific workflow]. The output gets [reviewed/used] in [specific way]." Without this, usage drifts.

Measure usage monthly. Pull the tool's usage data. If a tool isn't being used, either drive adoption or cancel.

Build a shared prompt library. What prompts work for what tasks. This compounds value across the team.

Hold a monthly retrospective. What did AI do well this month? What didn't work? What's worth trying next? Continuous improvement matters more than perfect initial setup.

These six practices implement the 70% on a real small business schedule.

When the 10-20-70 ratio doesn't apply

Two cases:

Very small AI investment (under $30/month). The fixed costs of process design and team training don't scale linearly. Below a threshold, the absolute hours required exceed the proportional value. For minimal AI use, simpler practices apply.

Single-user tools. When AI is used by one person (the owner) rather than a team, the team-training and process-redesign portions shrink dramatically. The ratio shifts toward 30-30-40.

For most small businesses with a real team and meaningful AI investment, the 10-20-70 ratio holds.

The honest takeaway

10% of AI investment on tools. 20% on integration and setup. 70% on people and processes. The ratio is well-established at enterprise scale and equally important at small business scale, even though it's almost always reversed.

The discipline that makes AI investments pay back is the people-and-process work. Tools without that work are paid-for waste. Tools with proportional people-and-process investment compound into real productivity gains.

For small businesses with tight budgets, the 70% is mostly time rather than cash. Pick an AI owner. Run weekly reviews during ramp. Document tool roles. Measure usage. Hold retrospectives. The hours invested in the 70% are the difference between AI as a cost and AI as a capability.