Model quality isn't what decides adoption
A sales VP asks why reps with access to a frontier model still copy notes into the same old template. The honest answer is rarely the model. It's AI workflow integration: whether the tool sits inside the systems people already use, or forces them to leave and come back.
Gallup's February 2026 survey of US employees found that those who strongly agree AI integrates well with their existing systems and processes are far more likely to be frequent users than those who don't.
A common reported reason employees give for not using AI more, per that same Gallup research, is simply preferring to keep doing work the way they already do it. That's a stated reason, not an observed behavior, so treat it as what employees say, not as proof of what actually slows them down.
Leadership impressions can also mislead: a vendor guide citing McKinsey research notes that C-suite estimates of how much employees actually use AI can differ sharply from what employees themselves report, which is one more reason to ask employees directly instead of guessing from the top.
We think workflow fit deserves the same upfront attention as licensing and security review, not a retrofit after adoption disappoints.
Sources: AI in the Workplace: What Separates Adopters and Holdouts, Measuring AI Adoption: Survey Questions for HR
What poor workflow integration looks like
Poor integration rarely looks like refusal. It looks like a rep opening a frontier tool, then a browser tab, then pasting the answer back into the CRM by hand.
- Tool switching: moving between the frontier tool and an older or built-in tool for the same task
- Copy-paste steps: retyping or pasting output between systems instead of it flowing automatically
- Missing data access: the frontier tool can't see the records it needs, so people fall back
- Unclear hand-offs: work stalls before it reaches the next person, team or system
A consulting firm's blog, Withum AI, walks through a scenario where a standalone AI tool that requires leaving the primary system adds three to five minutes of overhead to every single interaction, an example the firm frames as a workflow-design problem rather than a training gap.
The same vendor blog argues that when adoption stays flat despite training and setup, the real story is usually where the tool sits in the day's work, not whether people were taught to use it.
A separate vendor blog, amaiko, describes a familiar pattern: an employee leaves their CRM, opens a chatbot in another tab, pastes in context, and manually copies the output back once it's ready.
The Withum blog also notes that even when each tool works fine on its own, information can still stall between teams, so AI speeds up one step without fixing the slow hand-off that follows it.
We'd stop treating flat adoption numbers as a training gap by default; more often it's a workflow placement problem wearing a training costume.
Sources: The Deployment Problem Killing AI Adoption: Why AI Workflow Integration Must Live Where Work Happens - Withum AI, Why Do Only 25 Percent of Employees Regularly Use the AI Tools Their Company Is Already Paying For? | amaiko
Survey questions that reveal AI workflow integration barriers
Generic barrier questions ask people to check a box next to 'integration issues' and move on. Better AI workflow integration questions ask about each friction point separately, so you can track and fix them one by one.
| Friction type | What it signals | Sample question |
|---|---|---|
| Tool switching | Reps toggle between the frontier tool and a legacy or built-in tool for the same task | In a typical week, how often do you switch away from the frontier tool to finish a task in an older tool? |
| Copy-paste steps | Output is manually re-typed or pasted between systems instead of flowing automatically | For tasks where you use the frontier tool, how often do you manually copy or retype its output into another system? |
| Missing data access | The frontier tool can't see records or files employees need, so they fall back to the system that has them | How often does the frontier tool lack access to information you need, forcing you to use a different tool instead? |
| Unclear hand-offs | Work started in one tool or by one team stalls before reaching the next person or system | After you finish a task with the frontier tool, how clear is it who or what system should receive the result next? |
Aim these questions at individual contributors and team leads in the functions you're worried about, such as sales, support, or engineering, since the friction differs by role. Use a five-point frequency scale, from 'never' to 'every time,' for the closed questions below, and leave the last two open-text.
These are suggested prompts, not a validated instrument: adapt the wording to your own systems and roles, and treat the answers as what employees report feeling, not a measured account of time lost.
AI Adoption Score's own Barriers Survey asks this kind of question as one of nine predefined survey types, aimed at uncovering the real reasons a team isn't using AI more effectively.
Sources: AI in the Workplace: What Separates Adopters and Holdouts, The Deployment Problem Killing AI Adoption: Why AI Workflow Integration Must Live Where Work Happens - Withum AI, Why Do Only 25 Percent of Employees Regularly Use the AI Tools Their Company Is Already Paying For? | amaiko
Break the results out by team before you act
Perception gaps show up elsewhere too: in a survey of 1,200 global cybersecurity professionals conducted by Bitdefender and reported by TechTarget, far more managers than frontline practitioners said they had full visibility into AI tool use. The same kind of gap likely holds for workflow friction, which is one reason to ask each team separately instead of trusting a single company-wide number.
Break results out by team or function, not just by company average: a support team leaning on scripted macros faces different integration barriers than a sales team living inside a CRM.
AI Adoption Score's dashboards break results out by organization, team, and individual, which is the granularity this kind of question set needs.
The platform also maps each team against five maturity levels covering usage, skills, culture, impact, vision, and governance, so a workflow-integration gap shows up next to where a team actually stands.
We're skeptical of any single, company-wide adoption number; the teams hiding behind it usually have almost nothing in common.
Sources: Employees' shadow AI use is poorly monitored, survey finds | TechTarget
Turn reported friction into one fix at a time
If you change something, such as embedding the frontier tool directly inside the CRM instead of leaving it standalone, that's a direct fix for AI workflow integration. Run the same questions again afterward and compare the two rounds to see whether reported friction dropped.
Consider pairing these questions with ones about manager support and policy clarity: gaps in those areas may compound workflow friction, even though no finding here establishes that the two consistently move together.
Our view runs against the usual playbook: skip the all-hands retraining push and fix the one hand-off that's actually blocking a team's work first.
Stop guessing why adoption lags
You don't need another dashboard that says adoption is 'low.' You need to know whether the frontier tool fits where the work already happens, team by team, and what to move first. Ask the specific questions, read the answers as reported friction, and pick the one fix worth testing next quarter.
Frequently asked questions
Who should actually fill out this kind of workflow-integration survey?
Individual contributors and team leads in the functions you're evaluating, since friction varies by role and system. For mixed-role teams, tag responses by primary tool rather than by title, and keep responses anonymous when a function is small enough that answers could be traced to one person.
How often should we run these questions?
Establish a baseline first, then repeat the same questions after any deliberate workflow change, such as embedding a frontier tool inside the primary system instead of running it standalone. Comparing the two rounds shows whether reported friction moved.
Can these questions prove AI integration is hurting productivity?
No. They capture what employees say about friction and workflow fit, not a verified or causal productivity effect. If you want a productivity signal, compare the survey trend against independent usage logs or help-desk tickets over the same period.
Should we name the specific tools in the questions?
Yes, name the actual frontier tool and the actual legacy or built-in tool people compare it against. Generic wording like "AI tool" makes answers harder to act on, since you won't know which system people are falling back to.
