Every voiceon the team,one clear chord.

Chorda AI interviews a manager's colleagues in short voice conversations and synthesizes them into one report the manager can act on. People teams see the patterns across managers.

See how it works

A conversation asks what happened. A form asks for a score.

Chorda asks each colleague about real experiences, then connects their approved feedback into shared themes.

Example

“Small decisions wait for approval.”

“Escalation gets an answer the same day.”

Shared theme

Routine approvals are a bottleneck.

How a 360 runs

One cycle covers one manager. From the People team's side it is five steps, and the only one that takes effort is the first.

  1. Open a cycle

    Name the manager, approve a balanced set of colleagues, and set who receives the report. Everyone is told this before anything starts.

  2. Colleagues talk for about 15 minutes

    The interviewer asks about concrete experiences, strengths, and what would make working together better. Voice first, typing for the open-plan office. No accounts, no app.

  3. Each colleague approves their points

    The conversation becomes a short list of editable points. Colleagues change, remove or add any of them, choose how each one is shared, and submit. Nothing goes further without their approval.

  4. Chorda synthesizes

    It recognizes the same behaviour described in different words, weighs independent observations, keeps genuine differences, and runs privacy checks across the whole report before release.

  5. The manager gets one report. People sees the pattern.

    Strengths, development observations, differences in experience, and two or three behaviour experiments, written for the manager and their coach. People teams see the themes across managers, never individual submissions.

Three remarks in three vocabularies. One finding.

A summary repeats what people said. A synthesis works out what they were describing. Chorda recognizes the same behaviour in different wording, weighs the evidence, and keeps the differences in.

  • A vendor switch waited about two weeks for a yes, then happened the same day once it was escalated.

    Unprompted. A specific event.

  • Some people have stopped asking for approval on small things and just proceed.

    Unprompted. Describes a workaround.

  • Careful about decisions, rarely rushed. Appreciated when the stakes are high.

    Neutrally prompted. Experienced as a strength.
In the reportExample
Development observationIndependently recurring

Sign-off waits longer than the work

Requests that need your approval are described as waiting well beyond the time the work takes, then resolving quickly once raised directly. On small decisions, colleagues have started routing around the wait.

Difference in experienceContext-dependent

The same carefulness reads two ways

On decisions that are hard to undo, the deliberate pace is valued and colleagues describe wanting it. On small, reversible ones it is experienced as a bottleneck, and that is where the workarounds appear. The report keeps both, with the context that explains them.

Experiment to try

Set a response window

Agree a standard response time for sign-off requests and name the decisions people can take without one. Check in with two or three colleagues in six weeks.

Cues in words: "independently recurring" says how a finding is supported, "context-dependent" says its scope. The report never shows counts of who said what.

Why this is more than a tidied-up transcript

Same behaviour, different words

A waiting vendor, a quiet workaround and a compliment about carefulness are one pattern. Chorda connects them without flattening them.

Weighed, not counted

Points raised unprompted by colleagues who work with the manager in different ways weigh more than one person's repeated point, or what someone only confirmed when asked. Support shows as a cue in words, never a number.

Differences explained

When experiences disagree, the report names the context where it safely can: high-stakes decisions against small reversible ones, launch work against routine work. Never "some colleagues", never an average.

Checked as a whole

Before release, automatic checks confirm every finding is grounded in approved points and that the report as a whole does not point to a person. If they fail, the report is not released.

The observations and finding above are a fictional example. Reports never contain contributor names, exact wording, counts of who said what, or labels that could point to a person or a relationship.

A report written for the manager and their coach

Findings are written to be acted on. Each carries a cue in words, how it is supported or that it depends on context, and the whole report passes privacy checks before release. No scores, no quotes, nothing that points to who said what.

Every report ends with two or three behaviour experiments, each with signs of progress, so the next conversation between manager and coach has somewhere to start.

How you work with people

Summary report. Cycle closed.
Sample report

How to read this report. Findings describe experiences colleagues reported, in general terms. They are not verified facts or judgments of intent.

StrengthIndependently recurring

Context arrives with the ask

Colleagues broadly describe knowing why a piece of work matters before they start it, which makes trade-offs easier to make on their own.

Development observationIndependently recurring

Sign-off waits longer than the work

Requests that need your approval are described as waiting well beyond the time the work takes, then resolving quickly once raised directly. On small, reversible decisions, colleagues describe routing around the wait.

Difference in experienceContext-dependent

Feedback lands at different times

On launch work, your reaction is described as arriving right after the work, while it can still shape the next step. On routine work it is described as arriving at review time, when it is harder to act on.

The context is named because it does not point to anyone. Where it would, the report generalizes it.

Experiment to try

Set a response window

Agree a standard response time for sign-off requests, and name the decisions people can take without one.

Signs of progress at the six-week check-in

  • Fewer escalations to get a decision
  • People stop asking whether you have seen the request
Independently recurring is a support cue: raised unprompted by colleagues who work with you in different ways. Context-dependent is a scope cue: the experience differs by situation, and both experiences are kept. It says nothing about strength. Cues are words, never counts. The report never says how many people said something.

Each manager is one node. Leadership sees the constellation.

Colleagues' voices gather into one manager's report. Pull back, and that manager is one point among many. A theme that recurs across many of them is a reason to look at a shared process rather than at individuals, and the view shows how one shared moment landed in different teams.

  • Delegation
  • Micromanagement
  • Responsibility
  • Recognition
  • Manager or system?Example

    The same theme across a third of this year's cycles is a reason to look at the shared approval process before coaching twelve people.

    Sign-off waits12 of 40 cycles
    Context arrives with the ask23 of 40 cycles
    Feedback lands late9 of 40 cycles

    Counts are of manager cycles, never of the colleagues who spoke.

  • A shared momentExample

    One reorganization, raised in cycles across several teams, experienced two ways.

    The company view shows where each experience sits, without pointing at a team or a manager.

  • Same theme, next cycleExample

    "Sign-off waits", before and after a leadership programme on decision rights.

    This cycle7 of 38 cycles

    A change observed between cycles, and a reason to look closer. It is not proof the programme caused it.

A full 360 once or twice a year. Focused experiments in between.

Ongoing development instead of repeated surveys. The full conversation sets the direction. Short check-ins with two or three colleagues tell the manager whether an experiment is working, and each one knows which experiment was chosen and what the last check-in found. Chorda carries forward the experiment and the synthesized findings, not the conversations.

  1. Start of the cycle

    Full 360

    Colleagues talk for about 15 minutes each. The manager and coach choose two or three experiments from the report.

  2. Weeks 1 to 6

    Experiment

    One concrete change in behaviour, with signs of progress agreed in advance.

  3. Every 4 to 6 weeks

    Short check-in

    Two or three colleagues answer a few minutes of questions about that experiment only. Each check-in builds on what the last one found.

  4. Adjust

    Keep, adapt or close

    The manager sees what changed and picks the next experiment, or stays the course.

  5. Six to twelve months later

    Next full 360

    The full picture again, with last cycle's themes and the experiments tried alongside for comparison.

Trust is what makes the conversations good

Colleagues speak candidly when they know exactly what happens to their words. Chorda tells them before the first question, and keeps to it.

Every point approved by its author
Colleagues edit their own points and decide how each one may be shared before they submit. Nothing is saved in Chorda's systems before that moment.
What the manager and coach receive
One synthesized report, released only after automatic privacy checks across the whole document. Generalized where needed, or left out.
What nobody at the company receives
Individual submissions, recordings, transcripts, exact wording, counts of who said what, or a list of who has finished.
Honest about small teams
A report needs enough distinct contributors to be released, and that is a release rule, not an anonymity guarantee. Colleagues see the recognition risk explained before they decide what to share.

Consent per point, serious concerns, provider processing and the full participant experience are covered in the FAQ and on the approach page.

Built by the team behind Elsy and HeyAlum

Chorda is the third product from one team. Each turns a natural voice conversation into something useful afterwards.

An AI voice companion for older adults.

Everyday conversations help with the day ahead and, over time, become life stories. Memory gives each call continuity, so Elsy picks up where the last one left off.

helloelsy.com
Elsy's morning screen on a phone: a greeting by name and a reminders card describing the day ahead, with coffee with a friend at eleven.
Elsy's life stories screen: a chapter called Growing up in Maine, 1946 to 1960, opening with the story The ferry to Peaks Island.

Voice interviews and storytelling for alumni. In beta.

Alumni have a short call with their school's mascot. The call becomes stories the class can read, hear and add to, connected on a map of school moments.

heyalum.com
A HeyAlum call screen: the school's mascot listens and asks where the bus stop was, with send photo, mute mic and end call buttons.
A HeyAlum story page called The relay baton, told by an alumnus of the class of 2004, with a painting of sports day and a quote from the teller.

What carries over to Chorda

Voice conversations people are happy to finish, and the craft of turning them into structured memory rather than a transcript. Chorda brings both to feedback.

Many voices. One clear chord.

See how Chorda could work for your managers. Leave your details and we'll arrange a time.

Read our approach

Questions People teams ask

What do colleagues actually go through?

They open a link in the browser. The terms come first: who receives the report, what the company gets, how feedback is handled. Then about 15 minutes of conversation, by voice or by typing. The conversation becomes a short list of editable points, they adjust and approve them, and submit. The whole thing is described on the approach page.

How does consent work, point by point?

Every point carries identity protection by default: identifying cues are generalized, and a point is left out if it cannot be made safe. A colleague can switch protection off for a specific point to include it even if they could be recognized. Either way, the company never sees the point itself, only what the synthesis makes of it.

What if someone raises a serious concern?

Relevant feedback about serious behaviour follows the same consent and privacy rules as everything else. Chorda is not a formal reporting or emergency channel. Colleagues see your separate reporting route before they start, and nothing is forwarded automatically.

Where does the conversation go?

Nothing from an unfinished conversation or draft is saved in Chorda's systems. Close the page and it is gone. The voice and language providers that run the conversation process it under their own terms, which are shown to colleagues before they begin. Approved points are stored only after submit and deleted on a fixed schedule after the report is delivered.

Does the manager have to be involved in setting it up?

The manager proposes who should take part. The People team checks for balanced coverage, including direct reports, and opens the cycle. The manager does not interview themselves and does not see who has finished.

Can we see which manager a theme comes from?

No. The company view shows themes across cycles. It does not rank managers, cluster them by name, or expose any individual submission. The individual report belongs to the manager and their coach.

What does a check-in involve?

A few minutes with two or three colleagues, about one experiment only. It tells the manager whether the change is noticeable. Chorda remembers which experiment the manager chose and what earlier check-ins found, so the questions build on each other rather than starting from zero. It keeps the experiment and the synthesized findings, not the conversations. It is not a survey and not a monitoring tool, and it does not produce a score.

Which languages?

Conversations run in English. Tell us what your teams need when you book a demo.

Book a demo

See how Chorda could work for your team. Leave your details and we'll arrange a time.

Used only to respond to your request and sent to Chorda's team. Privacy

Demo request received.

We'll be in touch to arrange a time.