Use case · Performance reviews + 360s
The annual review that remembers the whole year.
Annual reviews shouldn't depend on what you remember from the last six weeks. Bonfiyah builds a longitudinal Speaker Insights profile per IC across every 1:1, every skip-level, every project meeting they're in. Speaker Themes surfaces patterns. Project Context drafts the synthesis. Anchored in source quotes, not paraphrase.
Features performance reviewers lean on
The Bonfiyah features built for review-cycle work.
Each one has a deep-dive page — click through if you want the full mechanics.
- Promise Tracker →
Every commitment from every 1:1, dated, kept-or-not — sourced to the quote. - Truth Layer →
When this quarter contradicts last quarter's narrative — flagged with confidence. - People Memory →
A profile per direct report — patterns, themes, what they own. - Speaker Themes →
Per-IC patterns over time — recurring topics, hedging signature, energy markers. - Team Dynamics →
9-box matrix on your team. Composition gaps before the calibration meeting. - Project Context →
Quarter-long IC briefing in 250-450 words across every 1:1 in the cycle. - Voice ID →
Persistent IC fingerprint — speaker carries across every meeting in the cycle. - Consent management →
Six-state consent — required for HR-sensitive recordings in many jurisdictions.
Four Pro AI surfaces, one IC profile.
Speaker Insights
Big Five trait profile + Pointillist vector + archetype, derived from every recording across the year. Stabilises as data accrues. The longitudinal read.
Speaker Themes
Recurring topics, hedging signature, energy markers, what the IC keeps bringing up. The pattern view that's invisible from any single 1:1.
AI Project Context
Year-end 250–450 word synthesis per IC: how they grew, what they shipped, where they struggled. The first draft of the formal review write-up.
Truth Layer
Flags drift between what the IC said in May vs October about clarity, motivation, or growth path. Sometimes evolution; sometimes a missed check-in. Surfaced in time to act.
Worked example
A year of 1:1s, then the review.
Across the year.
Twenty-six weekly 1:1s with one IC. Plus the four quarterly skip-levels her grand-boss sat in on. Plus the dozen project review meetings she presented in. Verbal consent at the start of each. The recordings accumulate into her Speaker Library entry — Bonfiyah's cross-recording Voice ID recognises her across all of them as the same speaker. She doesn't have to do anything; the manager's recording rhythm is what produces the data.
Mid-year — the calibration check.
The manager runs Speaker Themes on the IC at the six-month mark. The output surfaces patterns: she keeps bringing up "I'm worried I'm not visible enough to senior leadership"; her energy is highest when discussing systems-design work; she hedges noticeably on ambiguity ("I think we should... maybe... it depends"). Three concrete signals the manager wouldn't have surfaced from memory. The next 1:1 starts there — explicit conversation about visibility, intentional time on systems work, named conversation about hedging as a habit.
Q3 — Truth Layer flag.
The IC said in May "I have clarity on what I'm working toward — staff engineer by next year." In an October 1:1, after a project that didn't ship as planned, she said "I don't know where I'm headed; I might be plateauing." Truth Layer surfaces the inconsistency. The manager raises it directly: not as a gotcha, as a check-in. Turns out she'd never had the explicit conversation about whether the failed project changed her trajectory; the manager assumed it didn't, the IC assumed it did. Real conversation happens. The plan gets updated together. By the time review season arrives, the disconnect has been resolved on a real basis, not surfaced for the first time as a surprise.
Year-end — the review write-up.
AI Project Context generates the IC's year synthesis in ~400 words: trajectory, key shipped work, recurring themes, growth edges, where she's been most energised, where she's been least. Anchored throughout in source quotes — actual things she said in actual meetings. The manager's review write-up is an editorial pass over a structured first draft, not a from-scratch reconstruction of memory. Calibration meetings (where managers compare ratings on ICs across the team) become substantive — every manager has the same depth of evidence per IC, not "the one who comes most readily to mind."
For 360 cycles — multi-recorder triangulation.
If two peers and a skip-level also subscribe to Bonfiyah, their recordings of joint meetings with the IC contribute to her Speaker Library entry. The manager sees the full triangulated profile — Speaker Themes drawn from her interactions across the team, not just from 1:1s with the manager. Real 360 feedback without the survey friction; calibrated against actual conversations rather than against impressions.
Honest limits.
Consent + transparency are non-negotiable. Recording 1:1s without explicit IC consent is a category error. Use the verbal-consent prompt at recording start. Tell the IC explicitly what you record, why, who can access it, and what the year-end output looks like. ICs who object — respect that, take notes the old way for them.
Show the IC the model's output mid-year, not just at review. The difference between this being a coaching tool and a surveillance tool is whether the IC can see the AI summary and Speaker Themes view of themselves before review season. Transparency converts the practice from extractive to collaborative.
Speaker Insights is inference, not measurement. The longitudinal trait read sharpens with data but is still inference from conversational patterns. Use it as a starting point for review prep, not as the binding signal. The IC's actual shipped work, peer feedback, and 360 input weigh more than any model's read.
Bonfiyah is not a performance-management system. Lattice, 15Five, Culture Amp own the formal-review-form layer — goal tracking, OKRs, structured 360 surveys, calibration ranking. Bonfiyah is the conversation layer that feeds those tools. The structured review still gets written up there; Bonfiyah just makes that write-up much faster and much better-anchored.
Don't use this for performance-improvement plans without HR review. If a review is heading toward a PIP, the documentation has to clear specific legal and policy requirements. Speaker Insights outputs are useful internal evidence but should be reviewed by HR and counsel before they become formal performance documentation.
FAQ
Is recording 1:1s ethical?
With consent and the right framing, yes — and many ICs prefer it. Frame as: "I want to capture our 1:1s so I can do justice to your year at review time and so we both have a clean record." Most appreciate the rigor. The rare IC who declines, take notes the old way and log the preference.
Can the IC see what's in their Speaker Insights profile?
Speaker Insights is manager-side by default. The right practice for high-trust teams: share AI Summary + Speaker Themes outputs with the IC during the review, anchored on source quotes. Transparency is the difference between this being a coaching tool and a surveillance tool.
How does this compare to Lattice or 15Five?
Different layer. Those are the structured-form layer (goals, OKRs, 360 surveys, calibration). Bonfiyah is the conversation layer that feeds them. Use both. Bonfiyah's outputs become the input to the structured review write-up in Lattice.
Will this scare off ICs who feel surveilled?
If used carelessly, yes. The lift is transparency: tell the IC what's recorded, why, who can access, year-end output shape. Show them the AI summary mid-year. ICs who appreciate the rigor become advocates; ICs who object aren't being unreasonable, take notes the old way for them.
PIPs and performance-improvement documentation?
Speaker Insights outputs are useful internal evidence but should be reviewed by HR and counsel before they become formal performance documentation. Don't use Bonfiyah as the primary record of underperformance without clear policy alignment.
Want a review-prep workflow?
Tell us your team size and review cadence (annual, semi-annual, quarterly). We'll send a Bonfiyah configuration tuned for the cycle — including a 1:1 consent script and a transparency-posture template for the team.
No spam. We use ConvertKit. See our privacy policy.