The Real Test of Agentic AI Readiness: Can You Involve the Right People?
Skills Gap: Organizations struggle not with technical capability but with having the right people making decisions.
Design Involvement: Involving the right operational people as core team members is crucial for effective agent design.
Compounding Issues: Small context errors by AI agents compound over time, affecting customer satisfaction and trust.
Smaller Advantage: Smaller organizations can more easily implement agentic AI with fewer complexities than larger enterprises.
Infrastructure Needs: Successful AI deployment requires more than technology; it needs organizational capability and data quality.
When organizations pursue agentic AI, they lead with technology. They talk about agents, orchestration layers, multi-model architectures.
What they don't often mention: the people and what happens when things break in ways nobody predicted?
The problem isn't technical, but many leaders don't realize this until they're deep into deployment, watching customer satisfaction scores drop and quality metrics slide.
Keep Reading—and Keep Leading Smarter
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Have an account? Log In
The Agentic AI Skills Gap
Ask most executives about their agentic AI readiness, and they'll describe their AI engineers and data scientists. They'll detail their cloud infrastructure and model selection process.
What gets examined less is whether they've empowered the people who actually understand how agents will impact the business.
This is the real skills gap. Not technical capability, but organizational influence.
The question isn't "do we have AI engineers?" The question is "do we have the right people making decisions?"
In most organizations, the answer is no.
Only 14% of organizations have agentic solutions ready for deployment, despite 30% exploring options and 38% running pilots.
Designing AI Agents in a Vacuum
What happens when technical teams design agents without input from people doing the actual work? They oversimplify.
A process that looks clean on a flowchart contains invisible layers of judgment. The customer service rep who knows when to bend a policy. The operations manager who understands which exceptions matter. The analyst who reads between the lines of incomplete data.
These decisions run on feel and experience. They look like simple "if this, then that" logic on a decision tree.
But people in those roles understand something agents can't easily replicate; context.
They know that while a rule holds true in most cases, there are times when you have to make an exception. Times when you do something differently.
Francisco Marin, co-founder of Cognitive Talent Solutions, describes the challenge his team faced when building HR agents:
We observed that there were some agentic capabilities deployed for core HR processes like payroll or regulatory compliance. But we were missing that generation of use cases that affect people analytics, and especially those that were aligned with the network framework.
His team didn't just build technical solutions. They built systems that reflected how work actually happens, not how it appears on an org chart.
When you build agents without surfacing tacit knowledge first, you don't get immediate failures. You get delayed-fuse problems.
Compounding Quality Issues
The agent makes a decision that's technically correct but contextually wrong. On a single occasion, this isn't catastrophic.
But over time, these small quality dips compound.
You don't realize it until customer satisfaction surveys come in. Until you notice a pattern in the data. Until the damage has already spread through your systems.
By then, you're not just fixing an agent. You're rebuilding trust.
95% of generative AI pilots fail to deliver measurable ROI. The root cause isn't flawed models. Poor integration and misaligned priorities kill these initiatives.
Translation: organizations don't involve the right people in design.
Who Actually Needs to Be in the Room
The person in the role you're automating needs to be there.
Not as a consultant. Not as an afterthought. As a core part of the design team.
They need to help you steer the agent. Share their lived experience. Surface the "except when" knowledge that doesn't show up in documentation.
Dan George, who previously served as a CHRO before co-founding Cognitive Talent Solutions, explains the stakes.
I’ve run this process manually in the past. When I was a former CHRO, been in charge of people analytics at different organizations. I’ve had to come up with lists and send them to learning and development or other talent engagement teams.
His experience building agents taught him that automation without lived experience produces systems that look functional but fail in practice.
"Having it as an automated, authorized spot where an admin can go and just click, click, click, it just makes the process that much easier and gets us out of just selecting the typical mentors and mentees that we just always kind of go to right off the bat."
But getting there required involving the people who understood the nuances of mentorship matching, not just the technical requirements of data processing.
Lisa Jones, CEO of EyeMail, learned a similar lesson when implementing AI across her organization. "We asked each department: Where do you feel friction or repetition? Where could AI help you elevate creativity or clarity? What would you love to spend more time doing if AI handled the rest?"
This approach sparked ideas from marketing to operations and gave each team ownership over how AI showed up in their daily work.
The best way to capture lived experience is direct involvement. And that involvement needs to come with two things most organizations separate: upskilling and fair compensation.
You Can't Separate Agent Design From People Development
Most companies treat these as different initiatives. Agent design happens in IT. Upskilling happens in HR. Compensation happens in annual reviews.
This fragmentation kills adoption.
You have to bundle them to keep performance high. Agent design is a new skill. Understanding how agents amplify your productivity and influence matters more than understanding how to code them.
When you don't bundle these elements, you get one of two outcomes.
- AI sabotage - Cynical participation that undermines the initiative. People who know the system will fail and quietly ensure it does.
- Reputational damage - Word spreads that your organization extracts knowledge without investing in people. Your employer brand takes a hit you can't easily repair.
Jones describes how EyeMail avoided this trap.
We didn’t roll out formal training manuals. Instead, we created AI Curiosity Circles, weekly sessions where team members shared discoveries, tested prompts, and explored emotional impact. These weren’t just technical tutorials, they were storytelling labs.
With AI handling repetitive tasks, Jones' team leaned deeper into storytelling, innovation, and customer empathy.
"Curiosity became a core value," she said. "Team members felt more engaged, more valued, and more inspired to experiment."
You have to tie it all together so people see it as an investment in them, not a replacement of them.
Why Smaller Organizations Have the Advantage
Enterprise organizations have already automated heavily with RPA and predictive AI. They've built complex systems with multiple approval layers, compliance requirements, and legal constraints.
Adding agentic architecture to this complexity creates exponential failure points.
Smaller organizations can move faster. They have fewer processes to navigate, fewer decision makers to align and fewer compliance headaches.
The best agentic architecture isn't overly complex. The leaner you keep it, the less there is to break.
Agents work best when they can operate with clear boundaries and minimal dependencies. When they can make decisions without navigating Byzantine approval processes. When they can access data without reconciling fifteen different systems.
Smaller organizations have this by default. Enterprise organizations have to build it intentionally.
Nearly half of organizations cite searchability of data (48%) and reusability of data (47%) as challenges to their AI automation strategy. Your data isn't positioned to be consumed by agents that need business context.
The CEO-CISO Divide
CEOs are optimistic. 67.1% believe AI tools will help them make better cybersecurity decisions. 60.2% think their company is more prepared than peers to respond to AI threats.
CISOs are cautious. Only 58.6% share that confidence. Just 19.5% believe AI will strengthen cyber defenses, compared to 29.7% of CEOs.
This gap tells you everything about organizational readiness.
The people responsible for security, compliance, and operational risk see problems the c-suite doesn't. They understand that agents introduce new attack vectors. That multi-agent systems create debugging nightmares and governance models don't exist yet for autonomous decision-making at scale.
Only 14% of CISOs feel fully prepared to integrate AI into cybersecurity operations. More than 50% cite lack of support from CIOs and CEOs as their biggest challenge.
Whether your organization has aligned on what deployment actually requires shows up here first.
What the Divide Actually Means
When your CEO is optimistic and your CISO is worried, you have a strategy problem.
The business side sees opportunity. The operational side sees risk. And nobody's built the bridge between them.
This shows up in deployment decisions, resource allocation, and how you prioritize speed versus safety.
Marin and George built this bridge into their agent design process. When I asked about consent and transparency in their HR agents, George explained:
"We've got to make anything that we do with agentic AI have the right level of consent and ethical use of these automatic processes.
Marin added: "We had ongoing discussions with the team about some of the nuances of each use case separately. In the case of our talent retention agent, for example, we had the discussion of, does it make sense to provide these insights at aggregate level? Or does it make sense to do it at the individual level and notify the immediate supervisor?"
For most orgs, the job of building this bridge will be done by leadership, but it's not the CEO and CISO leading. Instead, the partnership between the CHRO (people and process) and CIO (data quality and best practices) that will create the happy medium for opportunity to be realized in a risk averse way.
Organizations that close this gap early move faster. They build governance frameworks before they need them. They invest in observability and monitoring and they create clear escalation paths for when agents make decisions outside their boundaries.
The Agentic AI Infrastructure
Agents need more than APIs and data pipelines. They need infrastructure most organizations don't have.
Identity and permissions management that works across multiple agents. Tool catalogs that let agents discover capabilities without hard-coding integrations. Policy enforcement that applies consistently regardless of which agent makes a decision.
And observability that lets you diagnose issues when multiple agents interact in unexpected ways.
This is no temporary problem either, with 65% of leaders cite agentic system complexity as their top barrier for two consecutive quarters.
It's important to remember that infrastructure isn't just technology, it's also organizational capability.
Jones learned this when building agentic workflows at EyeMail. Her team uses Miro AI to visualize customer journeys, identifying emotional touchpoints and friction zones. Their use of HubSpot AI provides behavioral insights and segmentation, helping them personalize onboarding flows and nurture sequences.
"We improved onboarding completion rates by 40%. Enabled flexible, modular journey designs for pharma, tech, and automotive clients. Delivered storytelling frameworks that meet clients where they are—emotionally and operationally," she said.
Data Quality Demands
Your customer appears as "Acme Corp" in the CRM, "Acme Corporation" in email systems, "ACME Inc." in contracts, and "Acme" in call transcripts.
Without entity resolution, agents fragment their understanding across multiple incomplete profiles.
73% of enterprise data leaders identify "data quality and completeness" as the primary barrier to AI success. This ranks above model accuracy, computing costs, and talent shortages.
You can't agent your way around bad data. You have to fix it first.
How to Deploy Agentic AI Successfully
Organizations that successfully deploy agentic systems do three things differently.
- Involve the people doing the work in agent design. Not as consultants, but as core team members. They compensate this work fairly and use it as an upskilling opportunity.
- Build governance before they scale. Clear policies for what agents can and cannot do. Escalation paths for edge cases. Observability that catches problems before they compound.
- Keep systems lean. They resist the temptation to build complex multi-agent orchestrations before proving single-agent value. They add complexity only when simpler approaches fail.
AI bridges the distance between vision and execution," says Jones. "What once required weeks of research, consultant input, and cross-functional alignment can now begin with a well-crafted prompt.
But the key is how you get there. Her team didn't just roll out formal training manuals. They created AI Curiosity Circles, weekly sessions where team members shared discoveries, tested prompts, and explored emotional impact.
"For me the future organization is a network that is powered by AI agents, where AI agents are released to deploy these micro-interventions at scale," Marin says. "There's a component of large multinationals acting as incubators of this new way of work, what we call a network-first future of work."
This doesn't make for good press releases. But it separates the 14% with deployed solutions from the 38% stuck in pilots.
The Question You Need to Answer Now
Agentic architecture isn't coming. It's here.
40% of enterprise applications will integrate task-specific AI agents by the end of 2026, up from less than 5% in 2025.
As adoption of agents increases, the question every business has to answer comes down to whether they can meet what the agent challenge demands. Not just technical capability, but organizational readiness as well. This means:
- The right people in the room.
- The infrastructure to support autonomous decisions.
- The governance to catch failures before they compound.
- The data quality to make context-aware choices.
- The willingness to invest in people as much as you invest in technology.
Most organizations will get this wrong. They'll treat agentic AI as a technical upgrade instead of an organizational redesign.
They'll be part of the 40% of agentic AI projects Gartner predicts will be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls.
Organizations that get it right will recognize that agentic architecture demands more than engineers. It demands the lived experience of the people whose work you're trying to amplify.
And they'll build systems that respect that reality.
