7 in 10 enterprises expected to abandon vendor-built agentic AI by 2028

Gartner forecasts that 70% of enterprises will abandon vendor-assisted agentic AI systems by 2028 as costs rise and customers remain dependent on outside engineering expertise. It also expects fewer than 20% of forward-deployed engineering engagements to convert recurring customer requirements into vendors' core-product features, while warning that conventional consulting may be marketed at premium rates as specialized AI delivery. The outlook highlights material execution, governance, knowledge-transfer and budget risks for enterprise AI deployments and AI-services providers.

Analysis

The investable implication for Gartner (IT) is less a direct earnings shock than a potential demand signal: enterprises are shifting evaluation criteria from AI pilot velocity toward total cost of ownership, maintainability, and governance. That favors platforms with standardized deployment, strong partner ecosystems, and reusable workflows over vendors whose revenue model depends on bespoke implementation. The near-term beneficiary set is likely incumbent IT-services firms—ACN, IBM, G, and CTSH—if customers reclassify work as conventional transformation spending, although this may come with lower pricing than premium AI-specialist engagements.

For AI software vendors, the key risk is that services-heavy bookings can inflate reported contract value without creating durable, high-margin recurring revenue. Over the next 1-3 quarters, investors should scrutinize professional-services growth versus subscription growth, implementation duration, renewal rates, and services gross margin; rising services mix alongside unchanged net retention would be an early warning of commercialization friction. IT itself may benefit from heightened demand for independent AI governance and vendor-selection research, but the thesis is contingent on enterprise technology budgets remaining intact rather than merely being reallocated.

Consensus may be too focused on whether agents work technically. The larger constraint is organizational ownership: deployments fail economically when customers cannot operate, audit, or modify them after launch. Over 6-18 months, this should compress multiples for services-dependent AI application vendors while rewarding software businesses that can demonstrate customer self-service adoption, low implementation labor per deployment, and partner-led delivery rather than vendor-headcount-led growth.

Key Decisions for Investors

• Maintain a tactical long IT versus a basket of high-multiple, services-intensive private-market AI proxies only if Gartner demonstrates accelerating contract-value growth or AI research/subscription attach in the next two quarterly reports; the public-market read-through is indirect, so do not chase on this report alone.
• Screen ACN, IBM, G, and CTSH during the next earnings cycle for AI bookings converted into recurring managed-services revenue rather than one-time consulting backlog. Favor ACN/IBM on evidence of expanding AI-related managed-services backlog; avoid adding if utilization weakens or pricing concessions offset volume.
• Use earnings as a risk-control checkpoint for enterprise AI software holdings: reduce exposure where professional-services revenue outgrows subscription revenue for two consecutive quarters, net retention falls, or management lengthens implementation timelines. Those metrics would validate a shift from scalable software economics to labor-intensive delivery.
• Watch 2026 enterprise IT-budget guidance and CIO survey data over the next 3-6 months. A broad capex slowdown would make the apparent shift toward lower-cost partner delivery a budget-cut signal rather than a durable competitive advantage for IT services.