At an RBD. roundtable of about 70 executives, roughly half scored one AI initiative that had met barriers as it scaled against the five components of the Capacity Assessment. Two in three placed the barrier in people or process. Half said leadership sign-off would move it first.
RBD. ran an executive roundtable at the Open Data Science Conference West on the barriers to scaling AI on 27 October 2026. About 70 executives attended, and roughly half answered the survey. Each respondent chose one initiative at their own company that had met barriers as it scaled, then answered three questions: where the main barrier sat, who owned the initiative, and what would move it to production first. Just over half of the respondents worked at companies with more than 10,000 employees, and business, data and technology roles were represented in equal numbers.
At a conference built for data scientists, no respondent named technology as the main barrier. Two in three named people or process. Asked what would move the initiative to production first, half chose leadership sign-off, more than data access and budget combined.
No respondent said the team lacked the skills or that the initiative should stop. Half said the initiative would move only after a decision from a leader above the team that built it.
What has to change in the organization, rather than in the model, for a stalled initiative to reach production?
RBD. defines absorption capacity as the maximum amount of operational, cognitive and cultural change an organization can absorb at one time without causing fragmentation, burnout or workflow collapse. The Capacity Assessment, the first step of Band 1, Right-Fit Technology, scores that capacity across five components. The roundtable used the same five components as its answer options for the main barrier, so each response placed one initiative on the framework.
The 4 respondents who chose “Other” were all in technology roles, 4 of the 7 in the sample, so the listed options missed what technology leaders said they needed. By barrier, leadership sign-off led for process (5 of 9), people (4 of 8) and data (3 of 4). Governance was the exception: 4 of its 5 respondents chose data access.
Of the 5 initiatives still owned by the pilot team, 4 were waiting on leadership sign-off.
Respondents at companies with more than 10,000 employees chose leadership sign-off most often, 7 of 14. At companies with fewer than 1,000 employees, data access and leadership sign-off tied at 3 of 8 each.
15 respondents described one tactic that had worked. The most common was a small, agreed first scope, expanded once the value was proven.
| What had worked (question E, 15 answers) | Count |
|---|---|
| Start with a small, agreed scope and expand on proven value | 4 |
| Set up guardrails, frameworks or a decision layer first | 3 |
| A specific tool or tool practice, such as Amazon Bedrock, VS Code with Copilot, or having AI tools check each other's answers | 3 |
| Visible leadership buy-in, transparency and cross-functional support | 2 |
| Demos and worked examples shown to management | 2 |
| Not specific enough to group | 1 |
For half of the initiatives in this room, the next step belonged to a leader above the team that built it.
RBD. will run the Capacity Assessment with your leadership team on one AI initiative of your choice and report which of the five components will limit it.