In 90 days we build a client-owned growth system on your servers — infrastructure, content engine, and ad agents — and hand it to your team. Then we walk away.
Lower-risk first yes: start with the M1 Foundation — ~BDT 6 lakh paid before Day 1, credits fully into the complete engagement, owned operating stack with live dashboards by Day 30 or M1 refunds. Conditional on your documented obligations in Annex A.
Revenue is real, growth is uneven, and you're paying rent on five SaaS tools that hold your customer data hostage. If you recognise three of the boxes below, this is built for you.
Past startup chaos, not yet a 50-person org. Revenue is real, growth is uneven.
Klaviyo, Hotjar, Chatwoot Cloud, Shopify Apps. Each holds your customer data hostage and raises rates yearly.
A few posts a week while competitors flood feeds. You know AI can fix this — you don't have time to build it.
You can scale spend but can't see what competitors are running, can't test fast enough, can't stop the manual grind.
Calls, DMs, comments hold real customer language. Today they're entirely unsearchable — they vanish the moment they end.
Twelve months from now, you want something that grows in value — not a recurring invoice for the same SaaS bundle.
Pre-traction operations. Founders without an internal team to operate the stack post-handover. Anyone who wants a long-term vendor relationship instead of a clean transfer of ownership.
Every customer record in your PostgreSQL. Every session in your PostHog. Every conversation in your Chatwoot. Inbox, lifecycle, analytics, ads agents — on your servers, controlled by your team.
Competitors can buy the same SaaS — or clone the same open-source tools. They cannot buy your data, your tagged conversations, your trained models, or the agents tuned on your customers.
Three months, three discrete systems, one fixed operator and one fixed clock. The same architecture adapts to your vertical — D2C, clinic, hospital, real estate, or premium education.
Self-hosted infrastructure, unified customer warehouse, universal inbox, hardened security baseline, Conversation Vault, and a chat-ops gateway. 27 deliverables shipped.
100+ human-approved AI assets per week — every word sounds like your customers, because it is. Brand-locked image LoRA, RAG over the Vault, daily insights to Chatwoot.
Meta + Google ad agents on your accounts, daily competitor intelligence, Hermes chat-ops across the whole stack, four live training sessions, full team handover.
Phiora's complete engagement is BDT 15 lakh, paid 40 / 30 / 30 across M1, M2, and M3 with each tranche tied to that milestone starting. The lower-risk first yes is the standalone M1 Foundation Engagement — approximately BDT 6 lakh, paid before Day 1, delivering the foundation build standalone on your server.
VPS + Coolify · MinIO · Uptime Kuma · PostgreSQL warehouse · Airbyte ETL · dbt unified-customer table · Chatwoot universal inbox · Metabase dashboards · PostHog · N8N workflows · Listmonk · Typebot · Cal.com · Plane · Hermes chat gateway · Twenty CRM where your vertical runs real deal pipelines · Conversation Vault (recordings + Bangla/English transcripts + DM/comment archive) · server-side conversion tracking · hardened security baseline · consent kit.
By Day 60 your feed runs at scale. Ads, social, lifecycle, reply — generated in your brand voice from real customer language, with a human review gate before anything publishes.
Calls, DMs, comments, and recordings enter the Vault. A vector index makes every conversation queryable. AI agents draft ads, social, lifecycle, and replies in your brand voice. Your reviewer ships the day's queue in <30 minutes. Nothing publishes unreviewed.
Brand-style-locked image LoRA · RAG layer over product catalog and Conversation Vault · cultural-moment content calendar · daily AI insights replacing weekly · lifecycle flows fully populated · category and educational hub pages live.
By Day 90 your Meta and Google ad agents operate on your accounts, your daily competitor feed is in Metabase, and your team has been trained on the whole stack. Day 91 is handover.
The ads loop is a deterministic n8n workflow with hard budget caps. Your chat coworker reads everything — but money only ever moves through the loop. Daily spend cap · 24-hour rollback · WhatsApp "STOP ADS NOW" kill-switch.
Meta ads agent — 42-tool self-hosted MCP on your own developer app · Google Ads write agent on Google's official client library · CAPI + offline conversions · Performance Max with segment-tagged feeds · ScrapeCreators competitor scrape · vector-classified ad inspiration feed · Hermes full-stack coworker · 4 live training sessions · Loom library · decision-tree cheat-sheets · 30-day async support.
Your stack answers you on WhatsApp, Telegram, Discord, or email. Reports, reminders, scheduled jobs — created by chatting, not configuring. Group chats are read-only by design; money never moves through chat.
"What was revenue last week?" "Which ad is dying?" "What did customers complain about yesterday?" Answers come from your warehouse, your inbox, your Vault.
"Every Monday 9am, send the weekend numbers to the team group." Reminders, scheduled reports, and simple workflows — created by chatting, not configuring.
Your marketing team @-mentions it for numbers, summaries, and drafts. In group chats it is read-only, by design — it informs, it never acts.
Write actions run only from approved founder DMs, behind an approval gate. Ad budgets never move through chat — that path stays in the capped n8n loop.
Telegram, Discord, email, and WhatsApp one-to-one run on official APIs. WhatsApp group participation runs on a dedicated separate number — never your business line — under a written acknowledgment. Version-pinned, upgraded deliberately, never auto-updated on your server.
Phiora has no delivered-client case studies yet. We won't invent any. What we have is a live, deterministic demo environment — Aura Academy, a fictional premium education business in Dhaka — running the full stack on Phiora's own VPS. You can message it, see the dashboards, and judge the architecture before you commit.
The demo is gated: ack-whitelisted numbers get the full agent; unknown numbers get one walkthrough-booking reply. We built it to be an honest demonstration under controlled conditions.
One operator, one build at a time. Your start date is booked, not shared. The same person runs scoping through handover.
30 minutes. WhatsApp call or Meet. We listen, you talk. If Phiora isn't a fit, we say so.
One week. Vertical-tailored deliverable list, fixed scope, fixed fee, your obligations in Annex A. Platform applications filed Day 1.
M1 → M2 → M3. Weekly Loom updates, monthly milestone review. One operator, sequential.
4 × 2-hour live sessions, full Loom library, laminated cheat-sheets, live restore demo, every credential rotated and transferred.
WhatsApp Q&A, capped at 2 hours/week. Included in the engagement fee.
Milestone-tied: 40% at M1 start, 30% at M2, 30% at M3. The complete engagement is anchored at BDT 15 lakh. The M1 entry engagement is approximately BDT 6 lakh, paid before Day 1, and credits fully into the complete engagement if you continue. Heavier verticals (hospital, real estate) sit at 18–20 lakh. Discussed on the discovery call, never quoted blind.
Your team receives the credentials, restore proof, training library, and operating responsibility at handover. The stack is built to keep running in your accounts without a recurring Phiora dependency.
VPS ≈ $40–80/month · AI tokens ≈ $50–300/month on your own keys · ad spend is yours. Zero recurring line items billed by Phiora. Your conversation data is processed on your keys under no-training API terms — the provider is named in your contract.
Every floor applies when the Annex A client obligations are met — access and credentials on time, the call recorder installed, a named daily reviewer, training attendance, and real consented customer photos where those ad formats run. We build the machines; you feed them.
An owned, operating stack with live dashboards by Day 30 — or M1 refunds.
Conditional on Annex A client obligations.
100+ human-approved AI assets per week by Day 60 — or the M2 fee converts to M3 credit.
Conditional on Annex A client obligations.
Kill-switch, spend cap, 24-hour rollback — live through Day 121. If guardrails fail, we eat the spend, capped at the M3 fee.
Voided by manual ad edits outside the agent. All in writing.
WhatsApp call or Meet. We listen, you talk. If we can't help, we'll say so. No pitch deck.
WhatsApp +880 1676 494107 Or find the in-page link in any section.Every customer record, recording, transcript, model, and agent runs on infrastructure you control. Phiora operates as a data processor during the engagement; on Day 91 all credentials are rotated and handed to your team, and Phiora retains no local copies. Your stack keeps running in your accounts because the asset is capital, not a subscription.
M1 is the foundation tranche — 40% of 15 lakh. Sold standalone at the same unit pricing, never at a discount, because the deliverables are the same ones we'd deliver in the complete engagement. The smaller commitment exists because the rest of the 90 days is the part most teams want to see running before committing the full budget.
If you continue into M2 and M3, the M1 fee credits fully into the complete engagement. If you don't, you still kept an owned operating foundation with live dashboards.
Named in the engagement letter, not buried in a contract:
Every milestone guarantee is conditional on these obligations. We build the machines; you feed them.
Correct. The 2026-07-16 decision record's greenlight criteria — pilot delivered ≤ 90 days, 2–3 KPI deltas expressible in taka, and the client's own team operating the stack for 30 days post-handover — have not been met yet. The first engagement is being booked now.
Until then, Aura Academy is the sales asset. It is a live, deterministic, fictional environment — full stack, dummy data, regenerating nightly. You can message it, watch the agent catch a planted problem, and judge the architecture before you commit. We refuse to fake a case study.
It is yours. Full credential handover, rotation of every secret Phiora touched, Phiora deletes local copies. After handover, the client is contractually accountable for incidents in the client's environment because the data and infrastructure are the client's.
Out-of-catalog requests are priced change orders or refused. The 2026-07-16 decision record is explicit: only the connector mix varies by vertical. We don't pivot mid-build for one client's bespoke idea. Scope is locked in week 1 of scoping and bound by the engagement letter.
Discovery → 30-minute WhatsApp call or Meet. If we agree on fit, we move into a one-week scoping window: vertical-tailored deliverable list, fixed fee, Annex A obligations, and platform applications (Google Ads developer token, Meta developer app) filed on Day 1. Then M1 starts on your booked start date — one operator, one build at a time, no parallel engagements.
You don't need a deck to book the call. Pre-filled message — your business, your vertical, your biggest question. We'll reply.
Message +880 1676 494107