As the AI boom rolls on, AI startups can stay ahead of the pack by specializing, measuring ROI, and focusing on risk management. In our next Part 2, we will address the 4th challenge – “Build a moat: managing platform risk and defensibility”.
In 2024, 45% of all VC funding went to AI startups, Gartner reports. In 2025, Pitchbook found that AI startups are attracting nearly 58% of all VC funding.
BCG reports that VC funding of AI startups totaled $101.5 billion in 2024, while in the first half of 2025 it had already reached $116 billion. “Notably, agentic and autonomous AI startups are at the forefront of this charge,” the report states. In its report, Gartner projects that tech buyers will spend $3 trillion on AI in the period from 2023 to 2027.
On the technological front, performance benchmarks are improving rapidly, according to the 2025 report by the Stanford Institute for Human-Centered AI (HAI). At the same time, operational expenses and hardware costs declined, while energy efficiency improved, the report states.
But the AI revolution is progressing unevenly and by stages.
In its recent report, “The State of AI in 2025,” McKinsey writes that just 39% of enterprises surveyed “attribute any level of EBIT impact to AI.” Most organizations that do report an impact state that less than 5% of enterprise EBIT is attributable to AI.
The exception is the top 6% of respondents, which McKinsey calls “high achievers.” Those organizations are using AI for “transformative innovation” to power growth and expansion, rather than focusing on cost savings and efficiency. The high achievers are the most likely to realize enterprise wide benefits from AI use.
Despite these advances, and the flood of capital, challenges remain. This article will examine three of the challenges facing AI startups, and some potential solutions.
The first challenge is the need for AI startups to differentiate themselves from the competition. There are two related problems, “commodification” and being “steamrolled” by LLM model developers.
By 2026, enterprise buyers will spend more on software with GenAI features than on software without them, according to Gartner.
By 2028, Gartner expects 33% of enterprise software to include agentic AI. The authors write that AI is becoming a required feature, not a difference maker. AI startups will need to differentiate themselves to avoid being treated as another low-value commodity.
Compounding the problem is the fact that “most AI startups will build their companies around an LLM developed by another firm” due to high development costs, as Oracle reports. And the LLM model developers are not standing still. In 2024, OpenAI’s Sam Altman notoriously promised to “steamroll” startups that build “little things” on its model.
The pace of steamrolling is accelerating in 2025. “In the past year, we’ve watched [companies like OpenAI and Anthropic] move aggressively up the stack, evolving from infrastructure providers into true product companies,” write Jaya Gupta and Ahsu Garg of Foundation Capital.
In this environment, specialization is key to avoid being “commodified” or wiped out by the next version of LLM platforms such as ChatGPT or Claude. Approaches to specialization include:
A second challenge for AI startups is delivering ROI on AI solutions. The key takeaway is that, at the enterprise level, customers are scrutinizing – and measuring – ROI. To stay relevant, AI startups need to focus attention there, too.
A 2025 global survey conducted by IBM of 2,000 CEOs of end-user organizations finds that AI adoption is progressing but hasn’t yet delivered the benefits expected. It found that only 25% of AI initiatives have delivered the expected ROI for their customers over the past several years, while just 16% have scaled across enterprise customers.
To address the issue, 65% of CEOs surveyed say their organizations are “leaning into AI use” based on expected ROI, and 68% say they have developed “clear metrics to measure innovation ROI effectively.” The CEOs remain optimistic that their AI investments will improve efficiency (85% agree) as well as growth and expansion (77% agree) by 2027.
To meet customer needs, AI startups should be able to provide both quantitative and qualitative measures of performance, according to Oracle. Quantitative measures include ROI and the AI solution’s ability to meet technical KPIs.
AI startups must also be able to measure qualitative results, per Oracle, such as the ability to process unfamiliar datasets, the relevance of results to the target audience, and the comprehensiveness of results.
The next three years will be critical, Gartner predicts: “Startup tech CEOs believe that AI will have a massive impact on their businesses over the next three years – followed by declining impact as AI is more broadly integrated and therefore less of a differentiator.” To prepare, AI startups should focus on providing “concrete value” and productivity gains to their clients.
A third challenge for AI startups falls under the general heading of “risk management” with special attention to regulatory compliance. Risk management should always begin with a comprehensive data strategy to protect sensitive data, enhance safety, and optimize customer experience.
Stanford HAI’s 2025 AI report finds that measurements of responsible AI (RAI) “remain rare” among model developers and others in the industry, despite improved benchmarks for factuality, safety, and other RAI metrics. Meanwhile, governments and international bodies are showing increased urgency on the topic of RAI, with new regulatory frameworks coming into force in the US, EU, OECD, and African Union.
In its 2025 report, Gartner finds that enterprise buyers are more concerned than ever about risk management and regulatory compliance. It predicts that, by 2026, companies that “prioritize obtaining industry-standard compliance certifications” will improve “client win” rates by 30% and achieve a 20% higher success rate in fundraising.
The report also finds that 75% of enterprise buyers have low risk tolerance. These buyers look for suppliers who can provide advice and guidance on technology decisions. AI startups able to meet compliance needs and serve as technology guides are best positioned for growth in the largest segment of the market.
In our next Part 2, we will address the 4th challenge. “Build a moat: managing platform risk and defensibility”.
The AI startup revolution is ongoing and requires continuous attention to risks and opportunities. If you would like to learn more, please contact us.
If you’re building or scaling an AI startup and want to:
you don’t have to do it alone.
If you are facing complex operational, financial, product, or data challenges and want a partner who can think strategically, design the right systems, and help you execute, connect with Steve Wullschleger and the Wull team. Schedule your consultation here: https://wull.com/contact/
Wull is a Silicon Valley–based management consulting firm led by founder and managing partner Steve Wullschleger, who has spent more than 20 years helping funded startups, mid-size companies, and large enterprises solve complex, cross-functional problems.
Wull acts as an integrated C-suite partner across strategy (CSO), finance (CFO), technology (CIO), operations (COO/Chief of Staff), and product and system design, bringing a “Think. Design. Do.” mindset to every engagement.
The team combines over 100 years of experience in management consulting for technology and services companies, with deep expertise in AI, SaaS, fintech, business intelligence, operations, and human-centered leadership. Their work spans building financial and operating infrastructure, designing data and BI systems, leading AI-powered growth initiatives, re-engineering processes, and driving results such as successful exits, 3x revenue growth, major efficiency gains, and large-scale fintech deployments.
Copyright ©️ 2025 by Stephen Wullschleger. All rights reserved.