AI coworker · lives in Slack

Not another AI tab.
A coworker in the thread.

Your whole company @mentions the same Franky, right in the Slack threads where work already happens. It remembers your org, cites every source, and gets used by the people who'd never open an AI tool.

Book a demo

95% of our own company uses Franky every week — and Sales adopts it faster than Engineering.

Multiplayer · Org-memory · Source-cited · Used by non-engineers

#ask-anything 204 members

mei.tan Marketing 3:42 PM

@franky do free-trial users get the onboarding webinar invite? and who owns that flow?

Franky App 3:42 PM

Pulled the decision from #growth-decisions:

Yes — trials with 5+ seats auto-get the webinar invite
Decided 2026-03-14; solo trials get the email sequence instead
Owner: @kane.lin — automation lives in the lifecycle doc

Want the link to the playbook?

Sources: Slack · Drive · Notion decision-log

Built for

B2B SaaS (payments & data API) Vertical SaaS Marketing & DX agencies Electronics distribution Finance · insurance · securities E-commerce platforms

If your team does the same customer work over and over — proposals, renewals, support replies — each time with different details, Franky is built for you.

"Isn't this already solved?"

Every AI uses the same models.
The difference is knowing your company.

Franky runs on Claude — anyone can get the same model. Wiring up MCPs yourself gets your 5 engineers a great private tool. The other 95% of the company never touches it, it forgets everything between chats, and it breaks the day someone changes a schema. Here's the difference in practice.

AI in a tab

One person learns.

A great answer, seen by exactly one person. Close the tab and it's gone. Teammates never see the question, the answer, or how they asked it.

AI in the thread

The whole channel learns.

Same question, asked where everyone works. The answer is public, cited, and remembered — the next person searches the thread instead of re-asking.

	ChatGPT / Claude	Claude + MCPs (build it yourself)	Another hire	Franky
In the thread, multiplayer	✗	single-player	✓	✓
Remembers your org over time	✗	per session	in 3 mo	✓
Adopted by non-engineers	✗	power users only	✓	✓
Cites sources for every answer	✗	✗	sometimes	✓
Someone maintains it for you	n/a	you do	✓	✓
Cost / month	$20	$50–500 + eng time	$8K+	$1–15K

Every AI can write SQL and call tools now. What matters is whether it lives inside your company — and remembers it.

Where Franky lives

In the thread — not a tab you switch to.

This is why non-engineers actually use it. No new app, no new dashboard — @mention Franky in the channel you're already in. Everyone in the thread sees the answer, and it's cited every time.

Slack — Crescendo Lab

cs-tickets-active

maya.f 10:48 AM

@franky Acme is asking about extending their legacy SMS plan. What was last quarter's grandfathering decision?

Franky App 10:48 AM

Found it in #pricing-committee 2026-Q1:

Legacy SMS grandfathered through 2026-09-30
30% discount on migration to WA bundle
Owner: kane.lin

↪ Pricing policy · Decision doc

Native Slack integration

@mention Franky in any channel, DM, or thread. Works on desktop and mobile. Every answer cites its source, links the doc, and names the owner.

Microsoft Teams — Marketing

Marketing

General

Q3 Campaigns

Brand

Analytics

Posts Files Notes

Krystal · 9:14 AM

@Franky — pull the May campaign report. Spend, conversion, NPS. One slide.

Franky Bot 9:14 AM

May "Summer Awakening":

Spend $24,810 (GA4 + Meta)
Conv 412 tickets · CTR 3.2%
NPS 8.4 (Typeform, n=128)

Drafted slide → Marketing/May-report.pptx

↪ GA4 · Meta Ads · Typeform · Ticket backend

Microsoft Teams support

Add Franky to any Teams channel or group chat. Cross-platform answers stay consistent — same memory, same citations, same trust loop.

Reads from

GitHub

Asana

BigQuery

Drive

Metabase

GA4

Meta Ads

Gemini

+ 200 SKILLs

Trust you can check

No source, no answer.

Every answer follows the same steps — and when Franky can't find a real source, it says so instead of guessing.

1 · You ask

@franky why did Acme's usage drop in May?

→

2 · Parallel retrieval

Queries every connected source at once

BigQuery · usage events Slack · #cs-acme thread Asana · open tickets Drive · QBR notes

→

3 · Cited answer

"Usage dropped 31% after their May 12 API migration — two failing webhooks, ticket open since May 14."

↪ 4 sources linked · owner named

Don't take Franky's word for it. Click the links and check.

Proof · Internal Dogfooding

This isn't a pilot demo.

Four months of real production usage inside Crescendo Lab. The numbers below are from our own BigQuery telemetry.

95%

Employee adoption
(92 of 138 in last 30 days)

4,708

Mentions in 30 days

Power users
(10+ mentions / week)

4 mo

Continuous production
since Jan 2026

Source: Cloud Logging · franky-franklin-production · 2026/4/13 — 5/13. B2B SaaS industry median adoption is 30–40%.

Heavy users · in their own words

What changes when Franky's in the workflow.

Paraphrased from interviews with Crescendo Lab's top 20 most-active Franky users. Roles preserved, names withheld.

BD proposals used to be 3–4 hours each. Now I close the loop in under one — Franky brings the 60%, I shape the rest.

BD Lead, Japan · 2026 Q2

I stopped waiting for the data team. I just pull it myself with two messages. The 'I'll get back to you' delay is gone.

Senior CSM, Taiwan · 2026 Q2

Franky guesses the customer's real ask 60–70% of the time. That's the slowest part of CS work — gone.

CS Operations Manager · 2026 Q2

Real customer · May 2026

An event-production team.
50 people. Five data platforms.

Marketing runs ticketed campaigns across GA4, Meta Ads, the ticket-sales backend, Looker Studio, and Typeform surveys. Putting one campaign report together used to take 4 hours across five tools. Production, marketing, and engineering each owned a piece — nobody saw the full picture.

We installed Franky in their Slack workspace and connected the five sources. Their marketing lead now asks in one line: "Pull this campaign's spend, conversion, and survey responses." One question. Five tools. Every answer cites its source.

Onboarded in 2 weeks · production usage since 2026-05-08 · token-based billing

Data platforms unified through one Slack mention

2 wks

From kickoff meeting to org-wide rollout

Employees across three departments

Before Franky

4 hours

Per campaign report · 3 people involved

GA4 Meta Ads Ticket backend Looker Studio Typeform

Manual cross-platform pulling

Spreadsheet copy-paste

Wait on Eng for the ticket export

Re-check totals before sharing

A spreadsheet nobody trusts on the first send.

After Franky

~8 minutes

Per campaign report · 1 person, in Slack

@franky Pull May "Summer Awakening" — spend, conv, NPS, top survey
                  themes.

Franky queries all 5 sources in parallel, cites every number.

Returns a ready-to-paste report skeleton.

Marketing lead polishes the framing, sends.

Same report. Trustworthy on the first send.

How Franky actually works

Franky gives you the 60%.
You bring the 40%.

Our heaviest users don't ask Franky and ship whatever comes back. They think the question through first, ask Franky, check the answer, then polish it for the customer. The slowest part — finding, pulling, assembling — disappears. The 40% that's left needs human judgment anyway, and the work still ships under your name.

STEP 01

You translate the ask

What the customer says is rarely what you ask. You turn it into one clear question — the same thing a senior teammate would do.

Human work

STEP 02

Franky finds + returns

Pulls from connected sources, cross-references, returns a structured intermediate answer with every source linked. No source, no claim.

Franky's job

STEP 03

You polish + ship

Tone, customer context, your company's voice — you add those in minutes. Same quality, with the slowest steps nearly gone.

Human work

The flywheel

Every question makes the next one better.

Every conversation becomes your company's own memory — who owns what, how your team talks, which decisions were made and why. None of it leaves your workspace.

That's why month 6 looks nothing like week 1. Generic tools start from zero every chat. Franky gets smarter the longer it runs — and the harder it gets to replace.

What Franky Does

Real jobs. Real teams. Asked every day.

Six jobs validated by Crescendo Lab's most active users — Customer Success, BD, Marketing, Engineering.

Customer Success

Draft a customer reply

"@franky customer X is asking about pricing migration — Pulls ticket history, finds the playbook, drafts the response."

Marketing / Product

Pull cohort metrics on demand

"@franky how many MAAC trials converted in March? — Queries BigQuery, returns the cohort breakdown with source SQL."

BD / Sales

Assemble a customer proposal

"@franky draft a proposal for Acme — they sell B2B logistics, evaluating CAAC. Pulls past similar deals, current pricing matrix, competitive playbook. Returns a 60% draft."

Customer Success Manager

Build a renewal context pack

"@franky pre-renewal context for Beta Co — Pulls usage trend, last 90 days of tickets, NPS, open feature requests. Flags the two risks I should address first."

Eng / SRE

Root-cause a prod incident

"@franky why did checkout error rate spike at 3pm? — Looks at deploys, Sentry, recent PRs. Names the commit."

Customer Support

Look up history before replying

"@franky what did we tell this customer about migration last quarter? — Finds the ticket history, the playbook applied, the decision owner. One reply, one consistent answer."

Who Franky is for

Three signs Franky will pay for itself.

We interviewed our heaviest users, and the pattern is consistent: Franky helps most when the work repeats but needs tailoring, customers are involved, and the knowledge is too much to memorize.

↻

Same job, different details — every week

BD writes a proposal per prospect. CSMs tailor a renewal per account. Support adjusts the answer per ticket. Same job every time, different details every time — exactly the 60% Franky takes off your plate.

Many customer touchpoints

Engage → propose → sign → onboard → support → renew. Every step needs company knowledge — and every handoff drops some of it, so someone has to go find it again.

SOT

Too much to remember — but it's all written down

Pricing rules, product edge cases, old decisions — scattered across docs, threads, and tickets. Nobody can hold it all in their head. But an AI that can read it answers correctly, with the source attached.

If this work is already painful enough that you'd pay to make it stop — and the answers exist somewhere in your company — Franky's first job is to go get them: fast, sourced, in Slack.

Pricing

Land cheap. Expand on usage.

Three tiers that map to a natural buyer journey: try → adopt → standardize.

Pilot

$1–2K/ month

1 team · capped usage

Less than a part-time intern.

30-day trial
Slack workspace install
Onboarding skill seed
Email support

Start a pilot

Most customers land here

Growth

$5–15K/ month

Org-wide · usage-based

Less than one entry-level hire — works across your whole org, 24/7.

Unlimited employees
Pay-per-token over baseline
Self-serve skill authoring
Shared Slack support channel

Enterprise

$30K+/ month

Private deploy · SLA · SOC 2

Replaces a team's worth of context-fetching.

Custom GCP deploy
Dedicated CSM
Custom SKILL development
SOC 2 & security review

Talk to sales

Every tier is transparent token pricing — AI usage at cost × 1.4 (an open 40% service fee that covers running the system, permissions, monitoring, and support). You see exactly where every dollar goes. POC trials run at cost.

FAQ

Common questions.

What buyers usually ask before installing Franky.

How long does it take to onboard Franky?

One afternoon for the pilot install — Slack app + 2–3 read-only integrations (GitHub, Drive, Metabase). Org-specific SKILLs ship over the following 2–4 weeks as your team's actual usage informs what to build.

What data does Franky read?

Only the sources you connect, with the same access rights as the person asking. Franky inherits Slack, GitHub, and Drive permissions at query time — it never sees more than the requester is already allowed to see.

What if Franky gives the wrong answer?

"No source, no answer" is built into the system prompt — every claim links back to the source it came from. A persona-drift detector flags when Franky reaches beyond its grounding. Honest take: single-domain queries are our strongest area. Cross-product, cross-domain queries are our weakest — we surface confidence, we don't hide it, and we're rebuilding this for Q3.

How fast does Franky respond?

Sub-second for cached SKILLs and direct lookups. 10s to a few minutes for cross-source synthesis (BD proposals, renewal context packs). Heavy users treat Franky as async — they fire the question, switch context, come back to a structured answer. Async-by-default in the UI ships Q3 2026.

Is Franky SOC 2 / GDPR compliant?

SOC 2 Type II is on the Enterprise roadmap for 2026 Q4. Enterprise deployments run entirely inside your own GCP project — data never leaves. Pilot and Growth tiers run on Crescendo Lab's managed GCP with regional data residency.

What happens to our data if we churn?

You take everything with you. Connected source data was never copied to our infra — Franky queries it live. The org memory built from your team's interactions is exportable as structured JSON on request, deleted from our systems within 30 days of cancellation. No lock-in by design.

How does pricing scale with usage?

Pilot ($1–2K) is flat-rate, 1 team. Growth ($5–15K) is org-wide with a baseline usage allowance plus pay-per-token over it — most customers settle around $8K/month. Enterprise is custom based on private deploy + SLA needs.

Can different teams run different sub-agents in one workspace?

Yes — each team can configure its own SKILL set, persona, and data sources under one Franky workspace. CS team's Franky and Eng team's Franky see different worlds; permissions inherited from Slack channels. One bill, many specialized agents.

Can we build our own SKILLs?

Yes — Growth and Enterprise tiers include self-serve SKILL authoring. Customers regularly contribute SKILLs back to the library (currently 200+). New orgs benefit from every SKILL that ships.

Stop asking people.
Start asking Franky.

Install Franky in Slack and turn your company's scattered knowledge into one place everyone can ask.

or email jin@cresclab.com

Not another AI tab.A coworker in the thread.

Every AI uses the same models.The difference is knowing your company.