ICONIQ just released their January 2026 “State of AI” report surveying ~300 software executives building AI products. Here are the 10 metrics that matter most.
1. Just About Even Mix Between Sales-Led and Product-Led Growth in AI B2B
Not a surprise per se but interesting to see.
2. Companies Are Using 3.1 Model Providers on Average (Up from 2.8)
Multi-model strategies are now standard. OpenAI remains most popular at 77%, but Gemini jumped to second place at 55% (from 43%). Anthropic/Claude sits at 51%. The pattern: OpenAI + 1-2 other providers.
3. High-Growth Companies Spend 57% of R&D Budget on AI vs. 38% Average
The gap is massive. Companies growing 100%+ YoY are allocating nearly 20 percentage points more of their R&D budget to AI development. At sub-$100M companies, AI allocation jumped from 25% to 45% year-over-year.
4. AI Gross Margins Hit 52% in 2026 (Up from 41% in 2024)
AI economics are improving fast. Companies projecting 2026 see average gross margins of 52%, up from 45% in 2025 and 41% in 2024. Companies with “balanced differentiation” (model + product) report the highest margins at 53%.
5. Few Companies Believe They Have Data Quality Ready for AI Agents
This is the quiet big deal. Your AI Agents won’t really perform if the data isn’t highly cleansed, segmented, etc. We’re still early in getting this right, and thus in getting AI Agents to work the way they can and should.
6. 40% of $500M+ Companies Are Actively Deploying Customer-Facing AI Agents
Larger companies lead in agentic deployment, not smaller ones. At $500M+ revenue: 40% actively deploying agents vs. 28% at sub-$100M companies. The hypothesis: operational maturity, workflow scale, and customer demand required to deploy agents safely.
7. 37% of Companies Plan to Change AI Pricing in the Next 12 Months; And Annual Commitments Most Common in Consumption and Outcome-Based Pricing
Monetization is unsettled. Currently: 58% use subscription/platform pricing, 35% consumption-based (up from 19%), 18% outcome-based (up from 2%). Of those changing pricing, 28% are switching to consumption-based, 15% to outcome-based.
8. Model Inference Becomes the Dominant Cost Driver at Scale
At pre-launch: talent costs are 32% of total spend, inference 20%. At scaling stage: talent drops to 26%, inference rises to 23%. Infrastructure and cloud remain constant at ~17% across all stages.
9. Inference Averages 23% of Revenue
Obviously different vendors are all over the place, but this seems a good rough yardstick to use. Where will that money come from? Smaller teams? Better products and more self-serve and PLG, vs. less sales-driven revenue? Lower marketing spend? Ultimately, everyone will have to figure this out.
10. Forward-Deployed Engineers Now Used for 32% of Customers (Up from 20% in 2024)
FDEs are becoming standard go-to-market infrastructure. 44% of companies use FDEs in a hybrid role bridging product and delivery. The median percentage of customers receiving FDE support is projected to hit 32% in 2026, up from 27% in 2025 and 20% in 2024.
The bottom line: AI leadership in 2026 is about execution, not experimentation. The companies winning are allocating more R&D budget, achieving better margins, deploying agents at scale, and building hybrid pricing models—while the model layer itself becomes increasingly commoditized.
