Course 3.12 · Curriculum 2026.1
Call intelligence, and what not to log
Turn answered calls into structured records at fractions of a cent each, and decide deliberately what never gets recorded at all. A live seven-day import logged 37 of 56 calls, skipped 19 on purpose, and reported the uncomfortable finding that only one of 37 external numbers matched a contact.
- Estimated effort
- 5 hr
- Lessons
- 42
- Modules
- 7
- Material
- 5 hr
Estimated effort is the sum of the lesson and lab times in this version of the course. It is not video runtime, which is lower.
Canadian dollars. One learner, lifetime access to this course version. Tax extra.
1 lesson free to preview
This course does not issue a certificate on its own. It counts toward the paths listed below, and those do.
01What you will learn
7 capabilities, stated as things you can do
Each one is what you should be able to do at the end, not what the course covers. If an outcome is not testable, it is not an outcome.
- Join three telephony feeds on one external id without creating duplicate call records.
- Write a noise policy, and defend excluding internal calls and unanswered inbound from unknown numbers.
- Request a speaker-labelled transcript against a strict schema, and validate it before writing.
- Cost a call in tokens, and compare it honestly against a per-seat conversation intelligence licence.
- Extract action items with an owner and a due date that land on the task board you already have.
- Report a match rate that embarrasses you, because burying it costs more later.
- Handle consent, retention, and the calls that must never reach a model at all.
02Curriculum
7 modules, 42 lessons
Every lesson shows its length and its type. A SCORM lesson looks like any other lesson, which is the point of ingesting packages rather than linking out to them.
7 modules · 42 lessons · 5 hr of material
Click-to-dial, a five-minute analytics poll, and a real-time per-rep socket for screen-pop, joined on one unique external id so a single call never becomes three records.
- VideoClick-to-dial, a five-minute poll, and a real-time socket: three feeds, one callPreview6 min
- VideoJoining on one unique external id, without creating duplicates6 min
- VideoWhy a five-minute poll exists alongside a real-time socket6 min
- VideoThe screen-pop: what the socket buys you that the poll cannot6 min
- ReadingWhat happens when two feeds disagree about the same call5 min
- LabLab: join three sample feeds on one external id15 min
An explicit policy, stated before a single call is logged. Internal calls and unanswered inbound from unknown numbers are noise, not data.
- VideoAn explicit noise policy, stated before a single call is logged6 min
- VideoExcluding internal calls, and why that is not obvious from the data alone6 min
- VideoUnanswered inbound from unknown numbers: robocalls and hang-ups6 min
- ReadingWriting the noise policy so a reviewer can check it against the code5 min
- LabLab: classify twenty sample calls against the noise policy15 min
- VideoWhat the noise policy would have logged if nobody wrote it5 min
Every answered call goes to a fast model with a strict JSON schema: speaker-labelled transcript, summary, category, sentiment, action items, amounts, people, site address.
- VideoEvery answered call’s audio, to a fast model with a JSON schema6 min
- VideoA speaker-labelled transcript, and why speaker labels are worth the extra step6 min
- VideoSummary, category and sentiment against a fixed schema6 min
- VideoAction items with an owner and a due date, landing on the board you already have6 min
- ReadingAmounts, people and site address: fields that need validation first5 min
- LabLab: run one call through the schema and validate the response15 min
A 45-second call costs about 1.4k input and 0.7k output tokens. A ten-minute call is roughly 20k tokens. Both are fractions of a cent.
- VideoA 45-second call: about 1.4k input and 0.7k output tokens6 min
- VideoA ten-minute call: roughly 20k tokens, fractions of a cent6 min
- VideoComparing that against a $100-plus per seat per month licence6 min
- ReadingWhat the licence buys that this does not, stated honestly5 min
- LabLab: cost a week of calls for your own team size15 min
- VideoWhere the real cost lives: not the model call5 min
A real seven-day import: 56 calls fetched, 37 logged, 19 skipped on purpose, across 8 reps.
- VideoA live seven-day import: 56 calls fetched, 37 logged, 19 skipped6 min
- VideoAcross 8 reps: what an uneven call volume tells you before you read a transcript6 min
- VideoReconciling the fetch count against the logged count6 min
- ReadingWhy 19 deliberately skipped calls is a feature of the design, not a gap5 min
- LabLab: run the noise policy against a week of sample call metadata15 min
- VideoWhat a second week of import looks like once the policy is tuned5 min
Only 1 of 37 external phone numbers existed on a contact or lead. Reported rather than buried, because burying it costs more later.
- VideoOnly 1 of 37 external numbers existed on a contact or lead6 min
- VideoReporting a match rate that embarrasses you6 min
- VideoWhy burying the number costs more later than showing it6 min
- ReadingWhat a 3% match rate actually implies about the CRM, not the call system5 min
- LabLab: compute your own match rate against sample records10 min
- VideoThe fix this gap actually points to, and the one it does not5 min
What a caller needs to know, how long a transcript should live, and the calls that get skipped before anything is sent anywhere.
- VideoConsent: what a caller needs to know, and when6 min
- VideoRetention: how long a transcript should live, and who decides6 min
- VideoThe calls that must never reach a model at all6 min
- ReadingChecking your local call recording law before running this on real calls5 min
- LabLab: write the retention and consent policy for your own rollout15 min
- QuizCheck: call intelligence7 min
03Before you start
What you need first
Assumed knowledge and setup
- Courses 3.4 and 3.6.
- A Developer Edition org. Sample recordings are supplied; you do not need a phone system.
- Check your local call recording law before running the labs on real calls. Module one covers it.
Courses that come first
- 3.4 · Calling a model from Apex safely8 hr · Advanced
- 3.6 · Validating model output before it touches a record4 hr · Advanced
Part of these paths
- The Full StackShareCo Certified Platform Engineer
04Who teaches it
Elliot Saha
Co-founder, Chief Technology Officer
Platform architecture, custom Apex and Lightning Web Components, AI systems, and the Academy platform itself.
05Reviews
No learner reviews yet
This course has not been taken by enough people to publish an honest rating, and we will not print invented quotes on a page that sells verification discipline. Here is what we can evidence instead.
- Assessment
- Lesson checks onlyThis course carries lesson checks. The certificate is issued by the path assessment, not by this course.
- Curriculum version
- 2026.1You enrol into a version. Content changes do not move you mid-course, and your required-lesson count is snapshotted at enrollment.
- Refunds
- 14 daysFull refund within 14 days if you have completed under a quarter of the required lessons. Stated here rather than in a footer.
Start 3.12
Enrol as an individual, or buy seats and assign this course to your team. Progress is reported per lesson, per module and per path.
- 5 hr estimated effort, labs included
- Lifetime access to the version you enrol in
- A ShareCo certificate is not a Salesforce certification