The Data Cloud question keeps coming up. Same with the deflection-rate gap nobody's resolving. Heres what Ive seen across 17 deployments.
Whats true about Data Cloud:
- Mandatory for Agentforce in production. You can sandbox without it.
- Right call when the AI needs to look things up from your own documents, or scan very large product catalogues.
- Wrong call when your use case is a basic customer service agent answering known questions. Native Service Cloud and Flows handles that fine.
- Many orgs are buying it because the rep said so, not because the use case needs it.
The deflection gap, plainly:
- Salesforce marketing: 80%
- What teams report in production: 40-75%, and often unclear vs hiring/training cost
- The fix: pin down what "deflection" means before you measure. A bot answering one question is not the same as resolving the whole interaction. Measure against the chatbot you already had, not the keynote.
4 patterns that decided whether our deployments shipped or stalled:
- Pilots beat programs. Narrow scope ships. Broad ambition stalls.
- Works best when your data is already clean. Connected trusted inputs decide more than any prompt or model choice.
- Internal agents are the safest place to start. Lower stakes, faster feedback, the team learns quicker.
- The agent knowing when NOT to act is a feature, not a failure. Thats where trust gets built.
Work out what it would actually cost to hire and train the people the AI is replacing. Thats your benchmark. Not the percentages on the marketing deck.
If an account exec is telling you "you need Data Cloud", the question to ask back is: what exact job is this for, what are we measuring against, and how long until it pays for itself? If they cant answer, they dont know what theyre selling you and Data Cloud is a sunk cost dressed up as enablement.
What are you actually seeing in your deployments? Wheres Data Cloud paying off, and where is it just adding cost?
Disclosure: I work on Agentforce delivery for Bluprintx, not selling licences. Mod here too.