One brain, every channel
Voice, WhatsApp, SMS, email and web chat share one memory. A caller who hangs up mid-order can finish on WhatsApp without repeating anything. Platform
Investors
300K conversations a month
DRING builds AI staff for phone and messaging and runs them on customers' own lines. Real callers, every day, in nine countries.
01 · The problem
The phone line breaks first
Every business runs on calls it cannot staff for. Inbound queues peak at the wrong hour, a first outbound attempt reaches only part of a list, and nobody listens back to what was said. We measured what that costs on real lines.
~40%of inbound calls are never picked up in a human-run operation we measured over six monthsLogistics, customer's own switch data
54 smedian wait before anyone speaks on a human-staffed line, across a 70-call random sampleB2B distribution, random sample
30 to 50%of a list is what the first attempt reaches, so without retries the rest of the demand the business already paid for goes coldAverage figures, vary by sector, country and list quality.
02 · The solution
We run the line, not the software
A customer brings a workflow and a phone number. DRING builds the AI staff member for it, puts it on the line and stays responsible for how it behaves once real callers are on the other end. The customer keeps the relationship and gets back a result it can act on.
-
Their number, their systems
The agent answers and calls on the customer's own number through their carrier, reads from their CRM or order system and writes back into it. Nothing moves to a new tool.
-
An agent built from their operation
Call recordings, documents and connected systems become an assembled agent that has to pass a quality bar before it takes a real call, and keeps improving from live calls after launch.
-
Every call ends as a decision
Each conversation closes as green, yellow or red with a next action: call back, hand to a person, do not call again. Outbound work stops being a list of recordings and becomes a list of decisions. See the outcome standard
03 · Product
What a customer actually buys
Six capabilities decide whether an AI staff member survives on a real line. Each one is a product page of its own; this is what it does in one breath.
Outbound that keeps trying, politely
Reach works a list inside calling windows and attempt caps, retries by result, and dials each contact at the hour they tend to answer. Reach
A record, not a recording
Every call leaves a structured outcome, a confidence score, topic and warning tags and a suggested next step, readable without playing the audio. Call analytics
Results land in the CRM
Connect selects who to call from HubSpot, Salesforce or Freshdesk and writes the outcome, summary and do-not-call flag back to the same record. Connect
Handover with context
When a call needs a person, the agent passes it to the team with what has already been said, instead of dropping the caller into a queue. How handover works
62 languages, tested per workflow
The platform can run 62 languages across voice and messaging. A workflow goes live only in the languages, accents and vocabulary tested for it. Speech core
04 · Technology
Not a wrapper around someone's API
The model is the cheapest part of this business. The parts that take years are the build loop, the voice stack, the telephony and the controls, and all four are ours.
Built for one company, proven before the first call
Each agent runs 1,000 to 10,000 simulated conversations drawn from the company's own calls and documents. Live calls then choose the next release. Inside the Agent Factory
Our own voice orchestration
Speech recognition, the reasoning model and the voice are swappable per agent without slowing the conversation, with routing and failover between providers. Custom voices, accent and pronunciation correction and voice A/B tests run on the same layer. How the speech core works
The switch is part of the product
We connect to the customer's phone system and the numbers its carrier provides, and run the routing ourselves. Voice is anonymized at the switch before it reaches any model, repeat spam callers are blocked there too, and failover and human routing are treated as operating requirements. Telephony layer
Tested every release, not declared once
Six internal standards, from prompt defense to data protection and operator override, are re-tested on every version. Prompt injection is met with four layers, two of them independent of model behavior, and every conversation passes an automatic security check. The controls
05 · Results
Measured on live lines, not in a lab
Platform figures cover all live workloads together. Sector figures come from single deployments and are shown without the customer's name. Period and scope definitions sit in the deck.
~72%of conversations reach a resolutionAll live workloads
81%lower operating cost at the upper end, against the human-only baseline the customer agreedSelected deployments
1 in 10reached outbound conversations qualifies as a lead, in the same band across three unrelated sectorsLogistics, fintech, consumer hardware
90%of a list reached at the upper end, with retries, calling windows and best-time dialingSelected outbound campaigns
Average figures across outbound campaigns; reach varies by sector, country and list quality.
Why retries are a product
The first attempt reaches a third to a half of a list. Retries take it to 85%.
Reach calls back by result, inside the hours the campaign allows, at the time each person tends to answer, and keeps going until the outcome is known. In one logistics campaign a lead qualified on the fifteenth attempt. Nobody schedules that call by hand.
What the customer sees at the end is not a call count but a short list of people worth calling back: in early fintech campaign tests, up to a quarter of answered conversations closed green.
06 · Growth
Volume compounds with every agent added
The curve is indexed and directional, not a monthly series. What carries it is parallel launches, more agents inside accounts that are already live, and partners opening markets we do not sell into directly.
- Deeper in each account
A second agent on an existing customer, a survey after support or a reminder after a sale, adds recurring volume on the integration that already exists.
- Wider by playbook
Each live workflow becomes a playbook the next customer in a similar operation can assess in days, in a new sector or a new country.
- Cheaper per conversation
Models and telephony are shared across tenants, so handling one more conversation costs less as the platform grows.
07 · Customers and model
Who pays, and for what
Mid-market and enterprise operators whose revenue depends on the phone: logistics, payments, refurbished electronics, health services, energy retail, hospitality, recruiting and B2B distribution. Agents run live today in 10 languages; the platform supports 62, and each new language is validated before launch.
Repeating questions
Logistics, e-commerce and refurbished electronics field the same order, delivery and repair questions all day. This is the work a standing agent absorbs first.
Conversations with rules
Payments, supply-chain finance and health services need identification, a clean handover and a record of what was said on every call.
Outbound first
Recruiting, real estate, energy and hospitality usually start with one outbound list, then add inbound cover once the first campaign has been worked.
- Subscription sized to usage
Customers pay monthly for voice hours or messages, concurrency and the number of agent types. Overage runs at the same unit rate, with no penalty tariff and no setup fee.
- Expansion inside the account
A second or third agent on the same customer starts from the same call history and integrations, so growth inside an account needs no new project.
- Partners open new accounts
GTM Partners, such as CX consultancies and system integrators, sell under the DRING name, and white-label partners under their own. Both bring their own customer relationships, adding accounts without a matching rise in direct sales cost. Partner program
Companies that have worked with DRING
Yıldız HoldingEasyCepCenoaNavlungoPentaTrukkerElithairMap ElektronikHabitat HealthOtelistKıta LojistikSkyTürkRemax
08 · Team
Built in Istanbul, founded in 2024
Six people own the company end to end across five remits, backed by product, AI labs, backend and client operations teams. A named person signs off every release that goes on a customer's line. Meet the whole team
- ACArda CezzarChief executive, commercial and partner agreements
- ABAyberk BozkuşProduct and operations, release sign-off
- CMCan MergenTechnology, platform and telephony
- OKDr. Osman KahramanAI labs, models and evaluation
- HKHande KocagözInternational business
- HYHakan YazganInternational business
Go deeper than this page
We are looking for growth-stage investors who know B2B SaaS and the contact center market, and who can open doors with resellers and technology partners. The deck carries the definitions behind every figure here; the data room carries the monthly series.