TL;DR
- Every tool looks great in the vendor's demo. The decision should be made on your calls, not on English demo recordings.
- The three questions that eliminate the most tools in practice: Hebrew transcription accuracy, analysis against your criteria instead of a generic template, and integration with your existing telephony and CRM.
- A good pilot is measured up front: define before it starts which numbers will be checked and what counts as success.
- True cost includes implementation, training and criteria maintenance, not just the license fee.
The conversation intelligence market is crowded, and every demo looks amazing. The trouble starts a month after signing, when the Hebrew transcripts turn out patchy, the reports generic, and no rep has opened the system since onboarding. These ten questions are designed to surface all of that before the signature, not after.
The ten questions
- 1. What is transcription accuracy on OUR calls? Not "supports Hebrew" but a number, measured on your real recordings.
- 2. Does the analysis run on our criteria? Your script, objections and regulation, not a generic checklist.
- 3. How does it connect to our existing systems? Telephony, CRM and workflow tools, replacing nothing.
- 4. How long to first value? Days, or a quarter-long implementation project.
- 5. What does the rep get? A tool that helps them sell, or just a system that reports on them.
- 6. Where does the data live and who can access it? Security, permissions and privacy compliance.
- 7. What happens in real time? Alerts and a live in-call copilot, or only after-the-fact analysis.
- 8. What does success look like for a customer like us? Numbers from a floor of your size and industry.
- 9. What is the full cost? License, implementation, training and ongoing criteria maintenance.
- 10. What does a pilot look like? On your calls, with success metrics defined in advance.
The question most Hebrew-speaking floors miss
It is question 1. A tool with brilliant analytics and weak Hebrew transcription is a weak tool: every score, alert and insight is built on the transcript. The only valid test is running the tool on five of your real recordings, including a noisy call and a heavy accent, and reading the transcript against the audio. Below 90% accuracy, analysis produces false confidence. A Hebrew-first tool like Saleso reaches over 95%.
Generic vs custom: the difference you feel after a month
Generic tools give everyone the same reports: talk ratio, keywords, sentiment. That is interesting for two weeks. Real value arrives when the system measures what matters to your business: your script's milestones, your market's objections, your regulator's disclosure duties. In the demo, ask to see how a custom criterion is defined and who maintains it over time.
A red flag
A vendor unwilling to run a pilot on your real calls, against success metrics you defined, is telling you something. Whoever is confident in their product is happy to be measured on it.
Running a pilot that actually measures
- Define 2 or 3 success metrics up front: transcription accuracy, correct objection detection, script adherence measurement.
- Run it on varied calls: strong and weak reps, short and long calls, different campaigns.
- Compare system output to human listening on a sample, to confirm the machine sees what an ear hears.
- Check rep and manager experience: do they open the system without being forced?
Frequently asked questions
What is the single most important question of the ten?
For a Hebrew-speaking team: transcription accuracy on your own calls. Everything else is built on the transcript, and a weak transcript turns it all into guesswork.
Is a big international tool better than a local one?
The right question is which works on your calls. International tools are strong in English and their own ecosystems; on spoken Hebrew most of them degrade significantly. The only comparison that counts runs on your recordings.
How long should the selection process take?
With this question list, two weeks to a month: a deep call with 2 or 3 vendors, a transcription test on your recordings, and a short metric-driven pilot with the leader.