Hear this workflow
An outbound follow-up checks the order and battery-health record, then opens the appropriate claim without calling the pending case resolved.
Case studies
What DRING agents do on real calls
Five live deployments, each measured on every call in its own period, from a one-hour campaign to a year in production.
Case studies are published anonymized by sector, without a customer name. The brands that gave written permission to be named are listed on the references page. Five deployments are documented below, each with its period stated; the list grows as more mature.
| Industry | Workflow | Headline number |
|---|---|---|
| E-commerce, refurbished electronics | Inbound support, outbound survey, store line | ~75% resolved without a handover |
| Finance, digital payments | Inbound support, survey pilot | 92.6% answered instantly |
| B2B distribution | Outbound dealer campaign | 17.3% hot-lead rate |
| Logistics, freight marketplace | Inbound driver hotline, 24/7 | 100% answered instantly |
| Logistics, international platform | Outbound lead qualification | 194 qualified leads |
Ecommerce, refurbished electronics
Most inbound calls, handled by AI
60 to 80%of inbound calls handled by AI after one year in production
~75%of those conversations resolved without a human handover
Up to 81%lower operating cost on this selected deployment
A large refurbished electronics marketplace in Turkey runs a high-transaction storefront: order status, warranty checks on graded devices, and returns, all at scale. Before DRING, inbound volume regularly outpaced the support team, with wait times growing around promotions and restocks, and there was no consistent way to check in with buyers after delivery or staff a dedicated branch line.
The agent
Three agents, three moments
An inbound support agent answers the main line, reads the order record, and resolves warranty and return questions on the call, refurbished-specific cases like a battery-health dispute included. An outbound agent calls buyers after delivery for a short satisfaction check, and a separate agent staffs a store and branch support line.
Hi, I bought a refurbished phone last week, order 48213, and the battery health shows lower than listed.
I have your order here. Refurbished units are graded on battery health at intake, yours is listed at 89%. If it's reading lower than that, I can open a warranty claim right now.
Yes please, it's showing 81%.
Claim opened. A replacement or refund option will be in your account within the hour.
ResolutionClosed on call
Handle time~40s avg
- Discover
Map the call reasons
The highest-volume reasons for calling, order status, warranty, returns, were mapped before any prompt was written.
- Build
Three agents, one order system
Inbound support, outbound survey and branch-line agents were built against the marketplace's order and CRM records.
- Test
Refurbished-specific edge cases
Simulated conversations covered grading disputes and mismatched battery-health readings before go-live.
- Launch
Gradual rollout
The agent took a share of inbound volume first, then moved to the full line once resolution held steady.
- Evolve
Tuning after handover
Once handover was enabled, average handle time fell to about 40 seconds, then the team tuned transfer rules so the agent stopped handing over calls it could resolve itself.
| Metric | Result | Note |
|---|---|---|
| Resolved without a human handover | ~75% | After one year in production, with contextual human handover when needed |
| Operating cost advantage | Up to 81% lower | Against the agreed human-only baseline for the same workflow |
| AI share of inbound volume | 60 to 80% | Grew to a stable band over the first year, with human support available for exceptions |
| Monthly run-rate | 17-18K calls | Current steady state, after one year |
| Volume processed | ~43,000 calls | Across one 2.5-month measurement window |
| Availability | 24/7, no queue | No wait regardless of call time |
| Average handle time | ~40 seconds | Since handover was enabled |
What we learned
About 75% of conversations closed without a human handover, with escalation still available when a case needed judgment. On this selected deployment, operating cost ran up to 81% lower than the agreed human-only baseline. The AI share grew steadily from launch to a stable 60 to 80% band over the first year, a trend worth watching rather than a single week's peak. Enabling handover cut average handle time to about 40 seconds, but also surfaced over-transferring as its own tuning problem, separate from resolution rate. The survey and branch-line agents cover moments the inbound line never sees.
Finance, digital payments
92.6% of calls answered instantly
479interactions in 30 days (417 support, 62 survey)
92.6%of inbound calls answered instantly
70.3%of dialogues reached a conclusion
96.9%positive sentiment
A digital payments platform handles transaction status, failed-payment and account questions by phone, the kind of call where every minute on hold reads as a warning sign. Support volume was growing and issues were time-sensitive, and the platform also wanted a reliable way to measure satisfaction after a support interaction, rather than relying on occasional manual outreach.
The agent
Live support, plus a survey pilot
The inbound support agent reads the account and transaction state, resolves what it can on the call, and hands over to a human with full context when a case needs judgment or a manual bank trace. A separate outbound agent runs a satisfaction survey pilot after support interactions.
Hi, a transfer I made yesterday still shows pending.
I can see it, it's sitting with the receiving bank right now, that typically clears within 24 hours. I'll text you the moment it settles.
Okay, and if it doesn't clear by tomorrow?
Then I'll open a trace with the bank directly, no action needed from you, just reply to that text if it's still pending after 24 hours.
OutcomeResolved by AI
SentimentPositive
- Discover
Map top call types
Pending transfers, failed payments and account questions were identified as the highest-volume, highest-urgency reasons for calling.
- Build
Wire into account systems
The agent was built against live transaction and account state, not a cached snapshot.
- Test
Simulate disputes
Failed-transaction and disputed-charge scenarios were run as simulated conversations before launch.
- Launch
Live on the support line
The inbound agent went live answering the main support line, with human handover always available.
- Evolve
Pilot the survey
An outbound satisfaction survey was run as a pilot on a smaller group before any decision to scale it.
| Metric | Result | Note |
|---|---|---|
| Interactions (30-day window) | 479 | 417 support, 62 survey |
| Calls answered instantly | 92.6% | No queue |
| Dialogues reaching a conclusion | 70.3% | 40 resolved by AI alone, 218 handed to a human cleanly with context |
| Sentiment | 96.9% positive | |
| Survey pilot recommend score | +54 | Average 8.4/10, n=13, small sample |
What we learned
In payments, instant pickup carries more weight than in most other lines: 92.6% answered instantly built trust before the agent addressed the actual issue. Resolution rate alone undersells the result, of the 70.3% of dialogues that reached a conclusion, most were handed to a human, and 218 of those handovers carried full context so the customer never repeated themselves. The survey pilot's +54 recommend score is encouraging, but n=13 is too small to call a trend, the next step is running it at a larger sample before drawing a conclusion.
Hear this workflow
A calm support exchange that checks the payment, explains what is known and records a follow-up without presenting a pending case as resolved.
B2B distribution
117 dealers reached in about one hour
117dealers targeted
92%dealers reached
98real conversations completed
17.3%hot-lead rate, 17 leads
A B2B technology distributor needed to run a campaign about a new offer across its dealer network, a relationship-heavy list where tone matters as much as reach. Calling 117 dealers by hand would have consumed roughly a full workday for a human rep, time the sales team did not have to spend on outreach that might not convert.
The agent
Outbound, then routed
An outbound campaign agent called through the dealer list, introduced the offer, and asked whether a sales rep should follow up with a quote. Interested dealers were flagged as hot leads, and the agent also noted cross-sell interest outside the original script.
Hi, this is DRING, calling on behalf of your distributor about the new component line, do you have two minutes?
Sure, go ahead.
We have early access pricing open this month. Would you like a rep to follow up with a quote?
Yes, we've actually been looking at exactly that, send someone my way.
OutcomeHot lead passed to sales
ReceptionPositive
- Discover
Segment the dealer list
The 117-dealer list and the campaign offer were reviewed before the call script was written.
- Build
Script the offer
The agent was built to introduce the offer, answer basic questions, and flag interest for sales.
- Test
Simulate dealer objections
Common pushback, timing, pricing, prior product versions, was run in simulation before launch.
- Launch
Run the full list
The campaign called the full 117-dealer list in a single session.
- Evolve
Capture the side signal
Cross-sell interest surfaced outside the original script is now something later campaigns are built to capture on purpose.
| Metric | Result | Note |
|---|---|---|
| Dealers targeted | 117 | |
| Dealers reached | 92% | |
| Real conversations completed | 98 | |
| Hot leads passed to sales | 17 | 17.3% of conversations |
| Campaign duration | About one hour | Versus roughly one human workday |
| Positive reception | 97.2% | |
| Cross-sell opportunities surfaced | 4 | Outside original campaign scope |
What we learned
A campaign that would have taken a rep roughly a workday to dial through finished in about an hour. Speed alone would not have mattered if the calls had landed badly, 97.2% positive reception is what made the speed usable, since dealer relationships are built over years. The 4 cross-sell opportunities were not something the campaign was built to find, a reminder that a conversational agent surfaces information beyond its script, worth capturing deliberately in the next campaign rather than treated as a footnote.
Hear this workflow
A short outbound exchange about stock, timing and the next conversation with an account manager.
Logistics, freight marketplace
A driver hotline that never sleeps
56calls in the first five days
100%answered instantly, no queue
87.5%were real dialogues
21.4%of calls came outside business hours
A freight marketplace connects drivers with loads, and drivers call about open jobs, rates, and their own profile at whatever hour they happen to be on the road. Before this line, calls outside business hours went unanswered, and that after-hours demand was effectively invisible: nobody could see how much of it there was, because there was no one there to pick up and log it.
The agent
One line, any hour
The inbound hotline agent answers questions about open loads and rates, captures driver profile details, and escalates price negotiations to the team with the context already attached, at 9 AM or 2 AM alike.
Calling about the Izmir to Ankara load, what's the rate?
That one's listed at a fixed rate. I can flag your counter-offer to the team if you'd take it for a bit more.
Yeah, flag it, I'd do it for more.
Got it, I'll flag it now, someone will follow up before the load closes. Anything else while I have you?
OutcomePrice negotiation escalated to team
TimeOutside business hours
- Discover
Map driver call reasons
Open loads, rate questions and profile updates were identified as the core reasons drivers call.
- Build
Wire into the load board
The agent was built against live load and rate data, not a static list.
- Test
Simulate negotiation calls
Rate pushback and counter-offer scenarios were run in simulation before launch.
- Launch
24/7 from day one
The line went live around the clock rather than during business hours only.
- Evolve
Watch the after-hours pattern
Early data on when drivers actually call is now informing how the line and its escalation rules get tuned next.
| Metric | Result | Note |
|---|---|---|
| Calls (first 5 days) | 56 | |
| Answered instantly | 100% | No queue |
| Real dialogues | 87.5% | |
| Positive outcome | 34.7% | 4 jobs accepted, 6 negotiations escalated, 6 driver profiles captured |
| Outside business hours | 21.4% | Some at 2 AM |
| Positive reception | 87.8% |
What we learned
21.4% of calls landed outside business hours, some at 2 AM, demand that existed before this line went live but had nowhere to go, coverage made it visible for the first time. Not every call needs a full resolution to count as a good outcome, escalating a price negotiation with the driver's context already attached was itself a useful result. Five days is a short window, and the ratios here, 87.5% real dialogues, 34.7% positive outcome, are a starting read rather than a settled trend, worth watching as volume grows.
Hear this workflow
A driver calls outside the usual office rhythm and gets the operational detail needed to continue.
Logistics, lead qualification
194 qualified leads from a dormant pipeline
6,804companies in scope
2,513companies contacted
1,602full conversations
194qualified leads handed to sales, 12.1%
An international logistics platform had a large pool of leads that had gone cold: companies once in conversation, never converted, never fully closed out either. Working that pool by hand would have meant pulling the sales team off active deals to make calls that might lead nowhere.
The agent
Qualify first, route second
An outbound qualification agent called into the dormant pool over five weeks, checked whether freight forwarding was still relevant, and asked a small set of qualifying questions. Companies that qualified were handed to sales with that context already attached.
Hi, we spoke with your team a while back about freight forwarding, still relevant for you?
Actually yes, we're shipping more internationally now.
Good to know. What's your rough monthly container volume, so I can route this to the right rep?
Around 40 a month, mostly ocean freight.
OutcomeQualified lead handed to sales
ReceptionConstructive
- Discover
Segment the pool
6,804 companies were identified as in scope from the dormant lead pool before dialing began.
- Build
Design the qualification script
A short set of qualifying questions was built to separate genuinely warm companies from dead ends.
- Test
Simulate cold reactions
Disinterested and skeptical responses were run in simulation so the agent could close out cleanly, not push.
- Launch
Run across five weeks
The campaign worked through the contact list in stages over five weeks rather than all at once.
- Evolve
Hand over with context
Qualified leads were routed to sales with the qualifying answers attached, so reps started from context, not a cold list.
| Metric | Result | Note |
|---|---|---|
| Calls placed | 9,666 | |
| Companies in scope | 6,804 | |
| Companies contacted | 2,513 | |
| Full conversations | 1,602 | |
| Qualified leads handed to sales | 194 | 12.1% of conversations |
| Constructive reception | 95.1% | Happy or neutral |
| Sales team time freed | 36 hours of conversation | Handled without occupying the sales team |
What we learned
95.1% constructive reception, happy or neutral, on a pool the team had written off as cold was the biggest surprise, dormant leads were mostly just leads nobody had called back. Reach was the actual bottleneck, not conversion, of 6,804 companies in scope, 2,513 were contacted across five weeks, so planning the next campaign means budgeting for how many companies can realistically be reached in a window. The number that mattered to sales was the qualification rate, 12.1% of conversations, not the raw call count of 9,666 calls.
Hear this workflow
A focused outbound qualification call that separates a real logistics need from a polite no.
From the DRING journal
Written from the front line.
Longer pieces on what we learned deploying voice agents, from health tourism and hospitality to call center operations. All articles
Sector
The Future of Healthcare Tourism Growth with AI Solutions
How international patient pipelines change when intake, qualification and follow-up run on voice agents in the patient's language.
Read SectorEnhancing Patient Engagement and Satisfaction in Health Travel with Realistic AI Agents
Why the voice has to sound like a teammate before a patient books a consultation abroad.
Read SectorDiversification of Customer Portfolio in Hospitality: A Thorough Look into Multi-Language Guest Relations
Diversifying a hotel's customer portfolio when the reception line speaks every guest's language.
Read SectorFootsteps of a Major Breakthrough: Application of AI Agents on Traditional Sectors
How AI agents are being applied across traditional, less digitized sectors.
Read Quality and testingReplicating Top Performers and the Role of Team Agents in Voice AI Solutions
On using team agents in voice AI to replicate the approach of top-performing staff.
Read Operating modelOvercoming Bottlenecks in Call Centers: Is Call Center Hiring an Issue of the Past?
A look at call center hiring bottlenecks and whether voice AI changes the equation.
ReadHow to read the numbers
Measured on every call, published by sector.
Each case reports the measures its workflow tracks, drawn from the same four definitions, so results can be compared across sectors and over time.
Handled by AI
Share of calls the agent took and worked itself, ending in a resolution or in a handover with the context attached. Deflection to a callback does not count. How many of those closed without a person is reported separately.
Answered instantly
Share of calls picked up on the first ring, with no queue. Measured on the full line, including nights and holidays.
Outcome rate
Share of conversations that ended in the outcome the campaign was built for: a lead qualified, a booking made, a survey completed.
Reception
Sentiment at the end of the call, labeled on every conversation and audited against human reviewers on a regular sample.
Also in production
Sectors with live agents, case studies in preparation.
Deployments that are live but not yet written up. Each gets a case study once it has enough production data to report with a stated period.
Anonymized by sector. Case studies carry the sector, not the customer name. Named brands are listed on the references page, and we arrange reference calls with each customer's consent. Ask for one on the contact page.