Sales reps spend just 40% of week selling, per Salesforce report. Generic AI tools often get numbers wrong, risking CRM adoption. Alpha Score 61/100.
Sales reps spend just 40% of their workweek actually selling, according to Salesforce's 2026 State of Sales report. The rest goes to research, planning, internal work and admin. That statistic comes from Salesforce itself – a survey of more than 4,000 sales professionals across 22 countries.
The finding matters because Salesforce's own AI tools, including Einstein, are supposed to fix exactly this problem: automate the non-selling work so reps can focus on deals. But a new Phocas survey of distribution businesses suggests the fix may not land. Half of the respondents said they are testing general-purpose AI tools like ChatGPT and Claude on their own, outside any business system.
The Phocas survey found that generic AI often gets the numbers wrong for distributors. The reason is data governance. ERP data is accurate at the transaction level, but it only becomes trustworthy for analysis once it has been layered and validated across departments. Generic AI tools skip that step. They connect to raw tables and produce answers that look correct but are not.
If Salesforce's customers replicate this pattern – using AI on ungoverned data – the AI-powered features in CRM could produce misleading forecasts, stale pipeline views and wrong margin calculations. That would erode trust in the platform and slow adoption. Salesforce's own report highlights the symptom: reps avoid CRM because typing notes into a desktop tool between meetings does not fit how they work. A bad AI answer would make the avoidance worse.
What would reduce the risk? Salesforce could invest in data governance layers that map definitions – margin, active customer, sale value – across the ERP, CRM and e-commerce stack. It could also partner with validation agents that check AI outputs for internal consistency before serving them to users. The Phocas survey notes that some IT teams already build such agents, but most users do not know they exist.
What would make the risk worse? Continued reliance on generic AI without distribution-specific context. The Phocas survey found that the best prompts are not usually specific to distribution businesses. If 'margin' means one thing in inventory and another in finance, a skilled prompt engineer and an unskilled one both get an answer, probably both wrong.
Salesforce's Alpha Score stands at 61 out of 100, labelled Moderate. The score reflects the gap between the company's AI ambitions and the data governance reality its customers face. The next catalyst to watch is whether Salesforce adds built-in validation and governed data connectors in its next Einstein update – or leaves that work to customers.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.