The Problem With AI Isn’t the Tech. It’s the Team

KEY TAKEAWAYS

• Roughly 70% of AI implementation challenges come from people and processes, not the technology itself.
• Teams stall in three predictable ways: Resisters who avoid it, the Willing but Lost who lack training, and Self-Proclaimed Experts who are overconfident.
• The most expensive people problems often sit at the top, with managers on the sidelines and executives out of touch with frontline reality.
• The fix requires leading from the top, training on real workflows for at least five hours, and building verification and human review into the system.

IN THIS ARTICLE

1. Three ways the team stalls AI
2. The people problems above the team
3. How to fix a people problem
4. How Sophia members fix the team side first
5. The bottom line

Every commercial real estate firm we talk to has AI tools. ChatGPT seats. A Copilot license bundled into Microsoft 365. A leasing chatbot. A lease abstraction trial someone signed up for last quarter. And at most of those firms, if you ask what AI has actually changed about how the business runs, the answer is a long pause.

It’s tempting to blame the tools. The model hallucinated. The integration was clunky. The vendor overpromised. Sometimes that’s true. But the research says it usually isn’t the main reason.

Boston Consulting Group’s study of AI value found that roughly 70% of the challenges companies hit when implementing AI come from people and processes. Technology accounts for about 20%. The algorithms themselves, the part that gets all the attention, account for about 10%. MIT’s 2025 GenAI Divide report reached a similar conclusion from a different direction: the gap between the few companies getting real value and everyone else came down less to model quality and more to how organizations learned to use it.

In other words, AI isn’t failing at your firm. Your rollout is failing at the human layer. And in our experience, that layer breaks down in three predictable ways.

Three ways the team stalls AI

1. The Resisters: “I’ll pass.”

Some people simply don’t use it. Some quietly work around it. A few actively undermine it.

This isn’t a fringe problem. In a 2026 survey of 2,400 knowledge workers and executives by Writer and Workplace Intelligence, 29% of employees admitted to sabotaging their company’s AI strategy, and among Gen Z the figure was 44%. Sabotage there includes refusing to use approved tools, using unapproved ones, and feeding company data into public AI. Of those who admitted it, 30% said they feared AI would take their job. Another 26% blamed a poorly executed company AI strategy.

In a CRE firm, resistance looks ordinary. The property accountant who keeps reconciling by hand. The leasing manager who turns off the automated responder because “residents want a real person.” The analyst who re-underwrites every number AI touched, just to be safe, and then tells the team it didn’t save any time.

Here’s the uncomfortable part: resistance is usually rational. If nobody has told your team what AI means for their jobs, they’ll fill in the blank themselves, and they’ll fill it in with the worst case.

2. The Willing but Lost: “I tried it. Now what?”

This is the biggest group at most firms. They aren’t against AI. They opened ChatGPT, asked it a generic question, got a generic answer, and quietly concluded it was a toy.

The data backs this up. Gallup’s 2026 workplace tracking shows three in four employees say their organization hasn’t communicated a clear plan for integrating AI. In Gallup’s 2025 research, even among people already using AI at work, only 16% strongly agreed that the AI tools their company provided were useful. BCG found that only 36% of employees feel adequately trained, and 18% of regular AI users had received no training at all.

The usage patterns are telling, too. A 2025 Aalto University study found that most people rarely prompted ChatGPT more than once per question. They asked, accepted the first answer, and moved on. That’s like hiring a talented analyst, giving them a one-sentence assignment, and never asking a follow-up question.

In CRE, the Willing but Lost are the asset managers who know AI could help with quarterly reporting but don’t know where to start, and the acquisitions associates who use it to rewrite emails because nobody showed them it could pull and compare T-12s.

3. The Self-Proclaimed Experts: “I’ve got this figured out.”

This group is the hardest to spot, because it looks like success. Someone picked up a few good prompts in 2024, became the office AI champion, and stopped learning. Their confidence went up. Their accuracy didn’t.

The research on this is striking:

• Confidence outruns skill. The Aalto University study found that people using ChatGPT consistently overestimated how well they performed, and the people who rated themselves as more AI literate were the most overconfident. As lead researcher Robin Welsch put it, higher AI literacy brings more overconfidence.
• Experts can feel faster while getting slower. In a 2025 randomized trial by the research nonprofit METR, experienced software developers using AI tools took 19% longer to finish real tasks. Afterward, they still believed AI had made them about 20% faster.
• AI can make wrong answers more convincing. In the well-known Harvard and BCG study of 758 consultants, those using GPT-4 on tasks just beyond what the model handled well were 19 percentage points less likely to reach the right answer than consultants working without AI. Their write-ups, meanwhile, came across as more polished and persuasive.

Now translate that to commercial real estate. A market comp set that sounds right but isn’t. A lease abstract that misses a co-tenancy clause. An investment committee memo that’s beautifully written and built on a bad rent assumption. Overconfident AI users don’t just make mistakes. They make mistakes that are harder for everyone else to catch.

They also create work for everyone else. Researchers at BetterUp Labs and the Stanford Social Media Lab coined the term “workslop” for AI-generated work that looks finished but isn’t. In their 2025 survey, 40% of U.S. desk workers said they’d received workslop in the past month. Each incident took nearly two hours to deal with, an invisible cost of about $186 per employee per month. About half of recipients said they now saw the sender as less capable and less reliable.

And even genuine experts go stale. The leading AI models change several times a year. A workflow that was best practice twelve months ago may now be slower, less accurate, or simply unnecessary.

The people problems above the team

The three groups above live on the front lines. But some of the most expensive people problems sit higher in the org chart.

• Managers who stay on the sidelines. Gallup’s research shows employees whose managers actively support AI are nearly twice as likely to use it frequently. As of May 2026, only 36% of employees at AI-integrating organizations strongly agreed their manager supports their team’s use of AI. BCG found just 25% of frontline workers say their leaders give them enough guidance.
• Strategy for show. In the 2026 Writer survey, 75% of executives admitted their company’s AI strategy is more for show than a real guide to outcomes. Teams can tell.
• A perception gap at the top. In Writer’s 2025 survey, 75% of executives believed their AI rollout had gone well, while only 45% of employees agreed.
• Nobody owns it. In that same 2025 survey, 71% of respondents said AI applications at their company were being built in silos. At a 75-person CRE sponsor, AI is often “owned” by whichever acquisitions associate seems most tech-savvy, on top of their actual job. MIT’s report made the same point at enterprise scale: line managers, not just a central AI team, need to drive AI into daily operations.

How to fix a people problem

If 70% of the challenge is people and process, then roughly 70% of your effort should go there too. That’s the ratio BCG found among the companies getting real value from AI. In practice, it comes down to five moves:

1. Lead from the top, out loud. The CEO and COO should use AI visibly, explain why the firm is investing in it, and be explicit about the attrition-not-layoffs approach.
2. Train on real workflows, and train enough. Five hours is the floor, not the goal. Use your firm’s own rent rolls, T-12s, leases, and investor reports.
3. Write it down. When your best AI user’s knowledge lives only in their head, it walks out the door with them. Document how AI is used for each workflow so it becomes a standard operating procedure, not a personal trick.
4. Build verification into the system. Cross-check AI outputs, keep a human approval step on anything that matters, and keep an audit trail.
5. Measure adoption and results by person. You can’t coach what you can’t see. Track who’s using AI, on what, and what it’s saving in hours and dollars.

How Sophia members fix the team side first

This is why Sophia CRE is built around Learn → Automate → Scale: your team learns AI, Sophia automates your slowest work, and you scale without adding headcount. The Learn step comes first for a reason.

• For the Resisters: Sophia’s approach is built on reducing team size through attrition, not cuts. Leadership joins live Q&A during training, so the message about what AI is for comes from the top.
• For the Willing but Lost: AI Workforce Training is six 90-minute interactive sessions tailored to your firm’s actual workflows, with homework and hands-on practice. Training produces skills documentation that serves as your firm’s AI SOPs, and sessions are recorded so management can bring new hires up to speed. Meanwhile, Sophia builds your custom AI workflows, with the first ones typically live within 30 to 45 days. They show up right in Microsoft Teams or Slack, where your people approve, reject, or modify the work. Nobody has to become a prompt expert to get value.
• For the Self-Proclaimed Experts: Sophia’s proprietary AI Synchronization Model™ has multiple AI models cross-check each other’s work, so the answer doesn’t rest on a single confident response. Every action lands in a full audit trail.
• For leadership: Mission Control, part of the Sophia platform, shows runtimes, approvals, and activity by person. A monthly scorecard compares results against your kickoff baseline in hours and dollars, so you know exactly where adoption is strong and where your team needs help.

The bottom line

The AI models your firm can access today are more capable than most teams know how to use. Buying another tool won’t close that gap. Getting your people from resistant, lost, or overconfident to genuinely skilled will.

The firms that win with AI won’t be the ones with the most licenses. They’ll be the ones whose teams learned how to use them, and never stopped learning.