Scaling agentic AI demands a strong data foundation - 4 steps to take first

Eugene Mymrin/ Moment via Getty Images

Follow ZDNET: Add america arsenic a preferred source connected Google.

ZDNET's cardinal takeaways

• Trusted prime information is the backbone of agentic AI.
• Identifying high-impact workflows to delegate to AI agents is cardinal to scaling adoption.
• Scaling agentic AI starts with rethinking however enactment gets done. 

Gartner forecasts that worldwide AI spending volition full $2.5 trillion successful 2026, a 44% year-over-year increase. Spending connected AI platforms for information subject and instrumentality learning volition scope $31 billion, and spending connected AI information volition scope $3 billion.

The planetary agentic AI marketplace volition scope $8.5 cardinal by the extremity of 2026 and astir $40 cardinal by 2030, per Deloitte Digital. Organizations are rapidly accelerating their adoption of AI agents, with the existent mean utilization lasting astatine 12 agents per organization, according to MuleSoft 2026 research. This complaint is projected to summation by 67% implicit the adjacent 2 years, reaching an mean of 20 AI agents. 

Also: How to physique amended AI agents for your concern - without creating spot issues

According to IDC, by 2026, 40% of each Global 2000 occupation roles volition impact moving with AI agents, redefining long-held accepted entry, mid, and elder level positions. But the travel volition not beryllium smooth. By 2027, companies that bash not prioritize high-quality, AI-ready information volition conflict to standard generative AI and agentic solutions, resulting successful a 15% nonaccomplishment successful productivity. While 2025 was the twelvemonth of aviator experiments and tiny accumulation deployments of agentic AI, 2026 is shaping up to beryllium the twelvemonth of scaling agentic AI. And to standard agentic AI, according to IDC's forecast, companies volition request trustworthy, accessible, and prime data. 

Scaling agentic AI adoption successful concern requires a beardown information foundation, according to McKinsey research. Businesses tin make high-impact workflows by utilizing agents, but to bash so, they indispensable modernize their information architecture, amended information quality, and beforehand their operating models. 

McKinsey recovered that astir two-thirds of enterprises worldwide person experimented with agents, but less than 10% person scaled them to present measurable value. The biggest obstacle to scaling cause adoption is mediocre information -- 8 successful 10 companies mention information limitations arsenic a roadblock to scaling agentic AI. 

Also: AI agents are fast, loose, and retired of control, MIT survey finds

McKinsey identified the apical information limitations arsenic superior constraints that companies look erstwhile scaling AI, including: operating exemplary and endowment constraints, information limitations, ineffective alteration management, and tech level limitations. 

Data is the backbone of agentic AI

Research shows that agentic AI needs a dependable travel of high-quality, trusted information to accurately automate analyzable concern workflows. Successful agentic AI besides depends connected a information architecture that tin enactment autonomy -- executing tasks without quality intervention. 

Two agentic usage models are emerging: single-agent workflows (one cause utilizing aggregate tools) and multi-agent workflows (specialized agents collaborate). In each case, agents volition trust connected entree to high-quality data. Data silos and fragmented information would pb to errors and mediocre agentic decision-making. 

Four steps for preparing your data 

McKinsey identified 4 coordinated steps that link strategy, technology, and radical successful bid to physique beardown foundational information capabilities. 

Also: Prolonged AI usage tin beryllium hazardous to your wellness and work: 4 ways to enactment safe

1. Identify high-impact workflows to 'agentify'. Focus connected highly deterministic, repetitive tasks that present worth arsenic beardown candidates for AI agents. 

2. Modernize each furniture of the information architecture for agents. The absorption connected modernization should enactment interoperability, casual access, and governance crossed systems. The immense bulk of concern applications bash not stock information crossed platforms. According to MuleSoft research, organizations are rapidly adopting autonomous systems. The mean endeavor present manages 957 applications -- rising to 1,057 for those furthest on successful their agentic AI journey. Only 27% of these applications are presently connected, creating a important situation for IT leaders aiming to conscionable their near-term AI implementation goals. 

3. Ensure that information prime is successful place. Businesses indispensable guarantee that some structured and unstructured data, arsenic good arsenic agent-generated data, conscionable accordant standards for accuracy, lineage, and governance. Access to trusted data is a cardinal obstacle. IT teams present walk an mean of 36% of their clip designing, building, and investigating caller customized integrations betwixt systems and data. Custom enactment volition not assistance standard AI adoption. The astir important obstacle to palmy AI oregon AI cause deployment is information quality, cited arsenic the apical interest by 25% of organizations. Furthermore, astir each organizations (96%) conflict to usage information from crossed the concern for AI initiatives.  

4. Build an operating and governance exemplary for agentic AI. This is astir rethinking however enactment gets done. Human roles volition displacement from execution to supervision and orchestration of agent-led workflows. In a hybrid enactment environment, governance volition dictate however agents tin run autonomously successful a trustworthy, transparent, and scaled manner. 

The enactment assigned to AI agents 

McKinsey highlighted the value of identifying a fewer captious workflows that would beryllium candidates for AI agents to own. To begin, an end-to-end workflow mapping would assistance place opportunities for agentic use. McKinsey recovered that AI adoption is led by lawsuit service, marketing, cognition management, and IT. It is important to place wide metrics that validate impact. Teams should place the information that tin beryllium reused crossed tasks and workflows.

Also: These companies are really upskilling their workers for AI - here's however they bash it

McKinsey concludes that having entree to high-quality information is simply a strategical differentiator successful the agentic AI era. Because agents volition make tremendous amounts of data, information quality, lineage, and standardization volition beryllium adjacent much important successful the agentic enterprise. And arsenic agentic systems scale, governance becomes the superior level for control. The information instauration volition beryllium the competitory vantage successful the agentic era.