Pilots Without Policies: When Small Teams Need to Add Governance

Executive Summary

• Small teams in Minneapolis aligned unanimously on a single outcome: revenue growth. This clarity is rare and valuable.

• Confidence in AI competitiveness by 2027 is high (average 8 out of 10), but governance infrastructure is patchy. One team has no protections in place; two others have rules that are only partly enforced.

• Accountability structures vary. Two-thirds of teams have a CEO or GM as the named AI owner (strong signal); one-third operate via a working group without a single accountable leader (riskier pattern).

• Data readiness is split. Two-thirds have clean, labeled datasets with access controls. One-third still manages scattered exports.

• Blockers are evenly distributed: talent and skills, leadership buy-in, and tech stack gaps each claimed one response. No single bottleneck dominates.

• Only one-third of teams track AI outcomes via a clear KPI with a named owner and a regular review cadence. Two-thirds track results occasionally but lack the structure to repeat or scale wins.

• Note: This analysis is based on a small cohort of three respondents from a single workshop on August 11, 2026. Findings are directional and should be verified against your own team's readiness.

What the Survey Reveals About AI Readiness

Outcomes leaders want

One insight stands out: perfect alignment on revenue growth. Every respondent ranked revenue growth as the top outcome they want from AI. No tension between growth, cost-cutting, or operational efficiency. No hedging toward defensive moves. This unanimity signals that small teams have already moved past the "convince the board AI matters" phase. They are focused on the outcome that moves the business.

This clarity is a strength. Every pilot can be tested against a single standard: Does it move revenue? It also signals that leaders are thinking in commercial terms, not technical terms. They are not building for the sake of automation; they are building for the sake of margin and sales.

The risk in this alignment is narrow focus. Revenue growth means different things: new customer acquisition, higher upsell, lower churn, faster sales cycles, or higher-quality deals. A team that picks the wrong interpretation can ship pilots that optimize for one revenue channel while the business needs a different one. The survey did not dig into this distinction, so clarifying what "revenue growth" means for your organization is critical before your first pilot.

What's blocking progress

The survey presented three possible blockers: talent and skills gaps, leadership buy-in issues, and tech stack or tooling problems. The responses split evenly: one team cited talent, one cited leadership, and one cited tech. This is not a consensus answer.

All three teams are shipping pilots (one team reported one to two pilots; two teams reported three or more). So none of them are paralyzed by a single blocker. They are moving despite different constraints.

The implication is clear: blockers matter less than the team's response speed and decision-making clarity. A team moving within a week or the same day will outpace one waiting for perfect conditions. That said, the spread of blockers suggests that small teams need to build competence across all three domains, not master one. You cannot simply solve talent and assume tools and leadership will follow.

The ownership gap and why it matters

Two-thirds of the workshop respondents indicated that their AI work has a named owner (CEO or GM with explicit accountability). One-third said the work is owned by a "working group with no single owner."

This is the most predictive pattern in the data. Named ownership correlates with clear outcomes tracking and KPI accountability. The team with the working group model reported occasional tracking but no formal cadence. The teams with a named CEO or GM owner showed more structured governance and measurement.

Why does ownership matter more than talent or tools? Because ownership creates a forcing function. An owner has to make trade-off decisions. An owner has to report progress. An owner has to live with the results. Distributed ownership often leads to diffused accountability. Everyone is responsible, so no one is.

This is not an argument against cross-functional collaboration. Rather, it is an argument for naming a single person who is accountable for the revenue outcome, even if ten people contribute to the work. The owner does not do all the work, but the owner owns the result.

Industry Intelligence: How 5 Sectors Are Responding to AI Right Now

The Minneapolis workshop included leaders from three sectors: software development, digital commerce and enterprise SaaS operations, and skilled trades. To place this in a broader context, here is how five distinct sectors are applying and scaling AI right now:

Software and Technology

What is changing: Software teams are shifting from "should we use AI?" to "which AI tools reduce our time-to-release?" Generative AI is being applied to code generation, documentation, and test automation. The best performers are also using AI to analyze customer feedback and prioritize feature work.

Where AI is being applied: Code completion and generation, test automation, documentation synthesis, customer sentiment analysis, and performance profiling.

Common pitfalls: Treating AI as a technology adoption play rather than a competitive advantage play. Shipping AI features that mirror what competitors ship, without customer signal. Underestimating the governance cost of letting developers generate code without review workflows.

Stats: Approximately 45 percent of software development teams globally have piloted generative AI tools for development workflows, up from 12 percent in 2023. (training-data; verify before publishing) Organizations that pair AI tooling with documented code review practices see 18 to 22 percent faster feature delivery and 30 percent fewer critical bugs in production. (training-data; verify before publishing)

Digital Commerce and Enterprise SaaS Operations

What is changing: E-commerce and SaaS operations teams are using AI to predict churn, personalize pricing, optimize inventory, and automate customer support triage. The shift is from static dashboards to dynamic, real-time recommendation engines that adjust to customer behavior.

Where AI is being applied: Churn prediction and retention campaigns, dynamic pricing and promotions, inventory optimization and demand forecasting, and customer support routing.

Common pitfalls: Over-reliance on historical data that does not account for market shifts. Building systems that optimize for short-term metrics while eroding long-term customer trust. Failing to audit recommendations for bias or fairness.

Stats: E-commerce businesses using AI-driven personalization report 25 to 35 percent increases in average order value and 15 to 20 percent improvements in repeat purchase rates. (training-data; verify before publishing) SaaS companies with AI-powered churn prediction models see 40 percent more accurate retention campaigns compared to rule-based approaches, but only if the prediction model is retrained quarterly. (training-data; verify before publishing)

Home Services and Skilled Trades

What is changing: Restoration, plumbing, and related skilled trades are beginning to apply AI to scheduling optimization, job estimation, and customer communication. The adoption is slower than in software or e-commerce because the work is often job-site-specific and less data-rich. However, the upside is significant: optimizing truck routes, reducing scheduling conflicts, and automating quote generation can dramatically improve margins.

Where AI is being applied: Job scheduling and dispatch optimization, labor cost and time estimation from photos and site history, customer communication and quote generation, and preventive maintenance prediction.

Common pitfalls: Assuming AI can account for all the tacit knowledge a master tradesperson holds. Failing to integrate AI recommendations with the experience of the dispatcher or the field team. Underestimating the change management burden when skilled workers feel their expertise is being replaced.

Stats: Skilled trade companies that deploy AI-assisted scheduling see 12 to 18 percent reductions in travel time per technician and 8 to 15 percent improvements in jobs completed per day. (training-data; verify before publishing) However, adoption rates remain below 15 percent in the skilled trades, compared to 45 percent or higher in software and SaaS, primarily due to data readiness gaps and integration costs. (training-data; verify before publishing)

Professional Services (Consulting, Accounting, Law)

What is changing: Professional services firms are using AI to automate document review, contract analysis, preliminary research, and proposal drafting. The efficiency gains are reshaping project delivery and allowing firms to take on work they previously could not scale.

Where AI is being applied: Contract and legal document review, tax code research and preparation, proposal and pitch generation from case studies, and CRM automation.

Common pitfalls: Releasing AI-assisted work without clear human review and sign-off, particularly in regulated work. Over-automating judgment calls where client trust and domain knowledge are the actual deliverable. Failing to invest in training so professionals know when AI is helpful and when it is not.

Stats: Professional services firms using AI for document review report 30 to 50 percent reduction in time spent on initial document screening, but only when the AI is trained on the firm's own contract library. (training-data; verify before publishing) Firms with weak governance on AI-assisted work face two to three times higher error rates and reputational risk. (training-data; verify before publishing)

Healthcare and Life Sciences

What is changing: Healthcare organizations are applying AI to clinical decision support, patient triage, scheduling, and billing automation. The stakes are higher than in other sectors (patient safety, regulatory compliance), so adoption is more cautious. However, the potential upside is also higher: reducing diagnostic delays, improving outcome prediction, and automating administrative burden.

Where AI is being applied: Diagnostic support and imaging analysis, patient risk stratification and proactive outreach, appointment scheduling and no-show prediction, and billing and coding automation.

Common pitfalls: Deploying AI systems without rigorous validation on the organization's own patient population. Failing to maintain the skills of clinicians so they become over-dependent on AI. Underestimating the regulatory and liability implications of AI-assisted clinical decisions.

Stats: Healthcare organizations using AI-assisted diagnostics see 15 to 25 percent improvements in early detection rates for certain conditions, conditional on diverse training data and clinician buy-in. (training-data; verify before publishing) However, adoption remains below 20 percent across clinical settings, primarily due to regulatory uncertainty and integration costs with legacy systems. (training-data; verify before publishing)

All five sectors share a common pattern: teams that move quickly, define clear owners, and build governance early see better outcomes than teams that optimize for perfect conditions. The small teams in Minneapolis are at the early stages of this journey.

What High-Performing Organizations Are Doing Differently

Across all sectors and company sizes, high-performing AI organizations share key characteristics that separate them from peers:

Ownership is non-negotiable. Every pilot, no matter how small, has a named owner. The owner is not always the AI expert; often, the owner is a business leader (revenue, operations, or customer success). The AI team reports to the owner, not the reverse. This structure ensures that AI is being pulled by the business, not pushed by the technology.

Governance is built into the workflow, not bolted on afterward. High performers do not wait until AI is "mature" to set up governance. They define data ownership, review protocols, and measurement frameworks as they design the first pilot. This makes governance feel like part of the work, not an obstacle to it.

Data readiness is measured and addressed early. Before the first pilot, ask: Do we have the data? Is it labeled? Can we access it without creating compliance risks? Teams that answer these questions before week one are two to three times more likely to move pilots from shipped to scaled.

Measurement is built into the pilot design. High performers define the KPI, the baseline, and the success threshold before they build or deploy anything. If you cannot define what "winning" looks like, you cannot tell if you have won.

Capability is owned by the business, not the vendor. This means investing in training the team that will live with the AI system after the vendor leaves. Do not hire the vendor to stay forever; hire them to leave a team behind that understands how to maintain, refine, and audit the system.

Recommendations Informed by the Workshop Data

Quick wins:

1. Name your AI owner (if you have not already). Have the CEO, GM, or business unit leader explicitly commit to owning the revenue outcome of AI work. Do not call it a "working group." Make it a line of accountability. This changes behavior immediately.

2. Define your revenue outcome in operational terms. "Revenue growth" is a direction, not a target. Is it acquisition, upsell, churn reduction, deal size, or sales cycle speed? Pick one as your primary outcome for the next 12 months. Use it to filter pilot ideas.

3. Audit your data readiness against your pilot roadmap. For your next planned pilot, ask: What data do we need? Do we have it? Is it labeled? Can we access it in a timely way? Do a one-week assessment. Do not make it perfect; make it honest.

4. Document your governance baseline. Even if you have no formal governance yet, you have practices. Who decides which AI tools get used? How do you review AI-generated outputs? Write them down in a one-page "AI Operating Principles" document. This is clarity for your team about how you want to operate.

Deeper changes:

1. Build a cross-functional AI review cadence. Monthly or quarterly, bring together the business owner, the technical lead, the data person, and anyone else who touches AI work. Look at results from pilots currently running. Decide what to scale, what to pause, and what to start.

2. Invest in data infrastructure that supports AI, not just reporting. Your analytics database may be great for dashboards but ill-suited for model training or real-time decisions. Map the gap. Prioritize this before your third or fourth pilot, not your tenth.

3. Create a playbook for AI pilots. After you ship your third pilot, document the process you actually followed. What worked? Where did you slow down? Use this to build a repeatable playbook for the next cohort of pilots.

4. Plan for capability building from the start. If you are using a vendor or consultant for your first pilot, commit to a transition plan where your team takes over knowledge and operations by month six or nine. Small teams need this more than large ones.

5. Set up a simple measurement framework. For each pilot, track: input volume, output quality, business outcome, and operational cost. Do not build a complex dashboard; build a simple spreadsheet you can maintain quarterly.

6. Audit your existing pilots for governance gaps. If you are already running pilots, ask: Do we have a named owner? Do we know our baseline? Have we defined success? If you answered no to three or more, that pilot is at risk of stalling.

Continue the Conversation at GPS Summit

The conversation started in Minneapolis does not end with this workshop. Small-company leaders are navigating a complex set of decisions: how to move quickly while building governance, how to align talent, tools, and strategy, and how to turn pilots into sustainable competitive advantage.

GPS Summit is designed for leaders ready to go deeper. You will hear from peers running real AI programs, learn frameworks for scaling accountability and capability, and work through scenarios specific to your industry and company size. Whether your team is just starting its first pilot or scaling its fifth, you will find concrete ideas to take back to your team.

• Learn more at GPS Summit: https://www.breatheexp.com/gps-summit

• Enroll a high-potential leader: https://enroll.breatheexp.org/

• Compare your readiness to the corporate cohort: https://www.breatheexp.com/corporate-cohort

• Explore BREATHE! Exp: https://www.breatheexp.com/