91.8% of marketers say AI delivered. Only 16.9% can name its owner
Most marketing teams are happy with their AI and cannot prove why. In Sprinklr's "Show Your Work" study, reported by PPC Land, 91.8% of practitioners running AI in live operation said results met or exceeded expectations. Only 27.2% could quantify the result precisely, and just 16.9% had a named owner for AI business outcomes.
That is a sentiment finding, not a performance finding. An agent without a baseline number and a named owner is a feeling, not a system, and feelings do not survive a budget review. The fix is cheap and it is not a tool: write down one outcome, one baseline and one owner for every agent you run.
What the survey actually found
Sprinklr surveyed 807 US practitioners between September 8 and 23, 2026, and published the results on October 28. A first caveat: the study is vendor-run and self-reported, and nobody audited it. The direction is more useful than the decimals. The numbers, as PPC Land reports them:
- 536 of the 807 respondents had AI in live operation.
- Of those 536, 91.8% said results met or exceeded expectations, and 2.8% said they fell short.
- Only 27.2% could quantify a result precisely. Another 58.4% could estimate it within a range.
- 16.9% had a named owner for AI business outcomes.
Put the first and last bullets side by side. Nearly nine in ten report success, and fewer than one in five can say who is accountable for it. Satisfaction is the only signal most of these teams have.
Why satisfaction without measurement is a risk
Satisfied teams do not retire failing agents. If the only evidence an agent works is that the team likes it, there is no moment at which the evidence turns against it. The agent keeps running, keeps consuming budget and attention, and keeps shaping customer-facing work, until someone outside the team asks for a number nobody has.
That question is coming. BCG's 2026 CMO survey, published June 15, found 96% of CMOs say AI is driving end-to-end transformation of their function, but only about a third have done the work. Its advice is blunt: convert current momentum into measurable, defensible results before CFOs lose patience. A team that can only say "we like it" will lose that conversation to a team that can say "cost per qualified meeting fell from X to Y."
Vendors will supply speed claims. You supply the baseline
The same week Sprinklr published the study, it opened a beta of six marketing agents. Its product page claims campaigns launch 10 to 15 times faster and content costs 40% less, with no stated methodology, and pricing was not disclosed. That is normal vendor behavior. It also shows who has to do the measuring: the buyer. No vendor will give you your own baseline, because it cannot see your pre-agent numbers, and a speed multiplier is not the same as a business result.
Ownership is the missing control
An owner is not the person who set the agent up. It is the person who is accountable for the number, and who has the authority to change, pause or retire the agent when the number moves the wrong way. With an owner, a bad month becomes a decision. Without one, it becomes a shared shrug.
In a small company the owner is often the founder or the head of marketing, and that is fine. What matters is that a name sits next to each agent. Sixteen percent is the share of teams in the survey that can do this, so doing it is an advantage.
A three-line spec for every agent
- Outcome. One metric the agent is accountable for, stated in business terms: qualified meetings booked, reservations kept, revenue recovered from abandoned carts. Not emails sent or tasks completed.
- Baseline. The same metric measured before the agent ran, or on a holdout group that does not get the agent. Without this, the outcome number has nothing to be compared with.
- Owner. One named person with the authority to pause or retire it, and a date on which they will review the number.
If you cannot fill in all three lines for an agent, that tells you its true status: an experiment. That is fine, as long as it is labeled as one and does not get funded as a system.
What to do about it
- List every agent and automated workflow you run in one sheet this week, including the ones a vendor switched on by default.
- Fill in outcome, baseline and owner for each. Mark the ones you cannot complete as experiments.
- Create a baseline where none exists, using a holdout group or a before-and-after window, before you expand any agent.
- Ask vendors for their method, not their multiplier, whenever a speed or cost claim drives a purchase.
- Review the sheet monthly and retire anything whose owner cannot show the number moving.