Kalonesis Harvest
A capture software for the moments where learning and development actually happens
It sits in the room, captures a declared set of moments, and sends what it captures to a cloud analysis platform that turns it into an evidence-based record of a person’s growth.
What it is
Kalonesis Harvest is software, connected to a cloud analysis and intelligence platform. There is nothing to buy for the standard deployment: it installs on hardware you already own. A phone, a tablet, a laptop, a desktop, the Android panel already mounted in a training room — whatever is in the room where the moment happens. For remote moments it runs inside the meeting tools the organisation already pays for. The platform turns what it captures into evidence. Nobody has to enter anything, attend anything extra, or fill in a form afterwards — which is the point, because that is exactly where the record breaks down today.
What it captures
Two channels, and they are not the same thing. The organisation declares the moments in the room. The person connects their own working life, source by source, the way they already connect it to every other assistant they use.
What it captures
- Selection interviews and the training moments around them
- Meetings
- Tutoring sessions
- Webinars
- Seminars
- Other moments an organisation declares as learning moments
- The moments where what was learned, or what was put into practice, is checked
What it does not do
This list carries the same weight as the one above it. A product that can read a working life is judged on who authorises each part of it, and on how it is switched off.
- It does not connect a source without the person authorising that source themselves
- It does not keep reading a source after the person revokes it
- It does not let a manager, an administrator or us connect anything on someone’s behalf
- It does not record a room without everyone in it being told
- It does not score behaviour, productivity, attitude or output
- It does not build a record of the other people in a captured room
- It does not decide anything about a person on its own
- It does not operate before the employer has completed what the law requires
And whatever the person connects
You already hand an assistant your inbox, your calendar, your drive, your contacts, your repositories — in some cases your whole machine. You did it because the answers got better the moment it could see the work. Development evidence is the same problem: a system that sees four sessions a quarter is guessing, and one that sees the work is not. So the second channel is as wide as the person authorises, over standard OAuth, one source at a time.
Calendar
Who someone actually spends their time with, which meetings recur, and how much of a first quarter went to onboarding rather than to the job. Time is the one resource nobody self-reports honestly.
Mail and chat
The questions a person keeps asking, who they escalate to, and how long a thing stayed unresolved. The same question asked three times in five weeks is a gap, and today it leaves no trace anywhere.
Documents and drive
What they produced, and how it changed after review. The gap between a first draft and an approved version is one of the few honest measures of what someone learned that month.
Code and reviews
Which review comments keep coming back on their pull requests, what they had to redo, and what stopped being flagged. A reviewer already knows this; nothing in the organisation records it.
Tickets, CRM and project tools
What they closed, what was reopened, what was reassigned. Applied capability, in the system where the work actually landed.
The learning platform you already run
What was assigned and completed, so the record can hold it next to what the person then demonstrated. Completion stops being the measure and becomes one input among several.
Meeting tools
Transcripts and recordings your organisation already produces in Teams, Meet or Zoom. In most companies the material exists; nothing is done with it after the summary.
More signal is not a nice-to-have — it is the whole difference
Every conclusion the platform draws carries a confidence value. With four sessions a quarter that confidence is low, the projections are wide, and the suggested next step is the generic one anybody could have written. With the work in view, the same machinery produces something narrow enough to act on: this specific gap, evidenced here, and the reason it matters for this role. The quality of what comes out is a function of what goes in, and we would rather say so than pretend a thin record is enough.
Wide because it is granted, not because it is taken
Each source is connected by the person, through the provider’s own OAuth consent screen, at a scope they can see. Each one can be revoked from the same place, and revoking it stops the flow and removes what it fed. Nothing is connected on someone’s behalf, and no manager can switch a source on for a person. What that changes is who holds the switch — not whether the law applies. Article 4 protects a collective interest, so one person’s consent has never been able to displace it: the procedure covers this channel too, and a source connected at work is still a tool from which the possibility of remote monitoring arises.
The platform behind it
The software captures; the cloud analysis and intelligence platform interprets. It turns a session into evidence — what was demonstrated, what was difficult, what changed since last time — and records how confident it is in each conclusion, along with the moment that conclusion came from.
The Learning Twin
The evidence accumulates into a continuously updated view of each person: what they know, what they are working towards, what a role still requires, and where the path appears to be heading. It begins at the selection interview. Confidence is stated alongside every observation and every projection, because a record built from real moments is still a model of a person and not the person.
How it works
The same loop the platform has always described, with capture at the start of it instead of data entry.
- 01
Capture
Be present in a declared moment and record what happens in it.
- 02
Understand
Turn the moment into evidence, with the confidence of each conclusion recorded beside it.
- 03
Decide
Use that evidence, the person’s objectives and the organisation’s policies to identify a useful next step.
- 04
Act
Propose it to a person, who approves, edits or discards it.
- 05
Improve
Record what followed, so the next conclusion is drawn on more than the last one was.
And then around again
An example, end to end
A new joiner is three weeks past their selection interview and two tutoring sessions in. Here is what actually happens, and where a person stands in the way.
- 01
Capture unit
Declared moment recorded
A tutoring session on the team’s deployment process. Everyone present was told before it began, and the session is on the list the organisation declared.
- 02
Cloud platform
Evidence extracted
The person explained the rollback procedure accurately and hesitated on the approval path. Both conclusions are stored with a confidence value and a pointer back to the moment they came from.
- 03
Learning Twin
Record updated
Their record now shows demonstrated knowledge in one area and an open gap in another, against the requirements of the role they were hired into.
- 04
Tutor or manager
Human review
A person sees the gap, the evidence behind it and the confidence attached, then decides what to do — including deciding the system read the moment wrong.
- 05
Cloud platform
Outcome recorded
What was proposed, what a person decided, and what happened next are all recorded, so the conclusion can be evaluated later rather than trusted now.
Illustrative view — not a live deployment.
Human oversight is a design constraint, not a setting
Nothing derived from a captured moment reaches a decision about a person without someone able to see the evidence, disagree with it, and act differently. Approval steps, permissions, retention and audit trails are part of how the product operates rather than options bolted on afterwards.
Where it runs
Software, installed on hardware you already own — a phone, a tablet, a laptop, a desktop, the Android panel already on the wall of a training room — and running inside the meeting tools you already pay for when the moment is remote. That is a decision rather than a stage we have not reached: selling a device would mean procurement, an installation date, one box per room and a works-council conversation about that box, all before anyone had seen a single piece of evidence. The analysis runs in the cloud in the first version. An on-premise option, for organisations that need the analysis itself to stay inside their own walls, is under evaluation for 2027 — and its shape is genuinely not settled yet, including whether it would involve dedicated hardware. We would rather say that than sketch an architecture we have not decided on.