The numbers Brief Episode map Takeaways Charts Quotes Clips Glossary Go next Transcript

Episode 91 · Enterprise · 46 min

Where the IoT world ends

A decade of industrial sensors left facility managers with dashboards and no fixes — Sensfix's claim is that the repair, not the reading, is the product. Text and vision models read the complaint, classify it and dispatch the nearest technician who already carries the part: a refrigerator that finds Kumar, five-star rated, two and a half kilometres away, and books him for Tuesday at 2pm. The company puts its obtainable market at $1.8 billion today and says first-visit fix rates move about 75%.

AI & machine learning SaaS & enterprise Founder journey

Balaji Renukumar

Founder, Sensfix · with Vishal Krishna

45:51

Said in this episode

▶ 23:07

$1.8B

Serviceable obtainable market for AI maintenance

The company's own early-stage estimate for a category it says it is creating, growing about 24% a year; he concedes very little market data exists and the long-run number could be tens or hundreds of billions.

▶ 27:30

~75%

Improvement in first-visit fix success

Company-reported customer metric, stated as the first-visit success rate going up by about 75% once the complaint arrives classified with the right skill and tool attached.

▶ 27:45

12–40%

Cut in annual maintenance contract spend

Stated as 12–14% typically and up to 30–40% at best, across annual maintenance contract and SLA expenditure; these are the company's figures, not audited results.

▶ 27:30

<1%

Repair failure rate claimed

Attributed to removing the communication gaps and lack of real-time multi-stakeholder visibility that produce avoidable repeat failures.

▶ 4:35

$3.2M

Grant funding raised

Figure cited by the host in the introduction; Renukumar separately describes a multi-million-dollar European Union grant won after rigorous due diligence because nobody else was doing this.

▶ 29:52

30–35

Core team, of 50–60 in total

Spread from Silicon Valley to South Korea, working round the clock — the scale at which he says delegation became the hard problem.

The brief

The argument in sixty seconds

Balaji Renukumar's claim is that a decade of industrial IoT solved the wrong half of the problem: factories, offices and apartment blocks are now thick with sensors producing data nobody acts on, and Sensfix, in his phrase, is the company that starts where the IoT world ends. It treats a maintenance complaint as an AI problem rather than a ticketing one — text classification, retrieval, extraction and summarisation to turn a rambling phone call into a real complaint, OCR on troubleshooting manuals and LCD meters, computer vision on camera feeds — and then dispatches automatically to the nearest technician who already carries the right tool and consumable. His argument for why this is overdue is human, not technical: the person taking your complaint is bombarded and stressed, and a bad mood produces bad service, while software has no mood. The commercial case runs through four operating metrics — first-visit fix success up about 75%, repair failure under 1%, mean time to repair falling, and 12–14% rising to 30–40% off annual maintenance spend — set against a category so new that no buyer has a budget line for it, which is why POCs are easy and the money takes two to six months. Sensfix built in Germany, Spain, Poland, the US and South Korea in parallel so it could never collapse into a services shop, went to Korea because LTE could not carry vision-heavy apps, and is now signing an MOU with IIT Kanpur on the premise that AI trained for Western facilities will not survive an Indian factory.

Worth your time if you are

Facility and plant managers sitting on sensor data nobody uses

Enterprise founders creating a category with no budget line

Operators running proof-of-concepts in four countries at once

Scientists weighing a lab career against a product company

OEMs thinking about service revenue after the sale

Episode map

Where the conversation travels

Every block is a chapter, coloured by what it's about. Click any of it to jump straight to that minute on YouTube.

AI & machine learningSaaS & enterpriseFounder journeySales, GTM & growthManufacturingProduct strategy

01Cold open: Kunigal to Kanpur to the Bay 0:00 Vishal introduces an AI company wiring together IoT, digital twins and AR/VR, and Balaji traces the route — Kannada-speaking Karnataka, IIT Kanpur, Germany, Korea, now California — arguing that the Indian upbringing that produced Silicon Valley's CEOs is itself an export. 02Sense and fix, after a COVID false start 2:57 Incorporated in the US in 2018, nearly killed by a pandemic that emptied the facilities it maintained, the company restarted growth in 2021 around what it calls a do-it-yourself digital maintenance platform. 03Two pillars, one superhuman dispatcher 4:51 Language models and computer vision have matured enough, he argues, to read every way humans report a fault and then act like a tireless dispatcher that picks the nearest technician with the right skill, tool and consumable in his car. 04Bad mood, bad service: the human bottleneck 7:28 Maintenance disappoints because a stressed, bombarded human team is between you and the fix — so Sensfix digitises the input instead, scanning manuals, LCD meters and camera feeds, and ends with the refrigerator that books its own repair. 05B2B now, B2B2C through the OEM 10:18 Enterprise software earns the right to go consumer, as Uber did from limousines; for now it is gated communities and a white-label path where an appliance ships with lifetime ticketing and a WhatsApp-style Sense Channel. 06Field-first, not tech-first 12:45 With European Union money and a strong AI team they could have gone tech-heavy, but customers wanted simple things solved — so the stack settled into five forms of text AI, OCR, and ground-level tricks like geofencing service visits. 07Why 5G took them to South Korea 15:54 The platform is packaged so enterprises can assemble their own maintenance stack, but vision-heavy apps choked on LTE — which is why a Korean government programme and one of the world's first full 5G footprints became a product decision. 08From streetlight switches to parallel markets 18:18 His first startup built heavy electrical switches for streetlights with a two-year warranty, and the maintenance burden birthed the app; rather than perfect one market, they co-developed with early customers in Germany, Spain, Poland, the US and Korea to avoid becoming a services company. 09Naming a market that does not exist 22:05 AI maintenance is a category with thin data: an obtainable market of about $1.8 billion growing roughly 24% a year, incumbents like ServiceNow, Zoho and IBM Maximo aimed at IT service, and a future in which hotlines, receptionists and email tickets disappear. 10POCs are easy, budgets are not 25:26 Because everyone lives the pain, pilots start quickly — but nobody has money set aside for a category that did not exist, so two to six months of proof precede the budget, after which the software becomes hard to give up. 11The four numbers a facility manager buys 26:56 First-visit fix success up about 75%, mean time to repair down, repair failure below 1%, and 12–14% to 30–40% off annual maintenance spend — plus a live chat channel that replaces the black hole of ticket-confirmation emails. 12Running the company on its own DNA 29:06 A core team of 30–35, 50–60 in all, works round the clock from California to Korea, and delegation is the current bottleneck — so the answer is to digitise and automate their own operations the way the product does customers'. 13The pitch, and a 1978 song about plumbing 31:44 To managers who installed sensors and then asked what to do with the data, the offer is workload down, satisfaction up, money saved and a demo sandbox in hours — after which Vishal detours to Frank Zappa's Flakes, whose broken plumbing made the same point in 1978. 14Indianising the AI with IIT Kanpur 35:05 India looked too futuristic a year or two ago and now looks fast-adopting, so Sensfix is running factory POCs around IIT Kanpur and finalising an MOU with the institute because AI built for Western facilities, he says, will not transfer. 15Scientist to businessman; the engineer wave 37:09 He does not miss patents or papers — violet to red, he has walked the spectrum — and defends the much-mocked engineering and medical college boom as the thing that seeded a generation now returning to India to build. 16Navodaya discipline, burnout and Mars 40:29 Boarding-school routine — sleep, food, discipline — is his defence against a burnout he admits he has started to brush, before the conversation closes on aerospace dreams, science fiction and too many Marvel movies.

Takeaways

Ideas to carry out of this hour

The sensor was never the product; the fix is

Facility and plant managers bought IoT expecting automation and got telemetry, and Renukumar's pitch is built on their disappointment: they installed sensors, collected data, and are now asking what to do with it. Sensfix positions itself as starting where the IoT world ends — harvesting whatever data and predictions already exist and converting them into a scheduled service action. The value it claims is not insight but a technician at the door with the right part.

▶ 31:59 ▶ 32:13 ▶ 32:29

Maintenance fails on mood, not on technology

His diagnosis of bad service is deliberately human. Between the complaint and the repair sits a team of people who are busy, bombarded and stressed, and someone attending to your problem in a bad mood will not serve you well. Software, he argues, has now matured enough to absorb the same messy inputs without emotion — which is why the first thing the product digitises is not the machine but the complaint about it.

▶ 7:28 ▶ 7:42 ▶ 7:57

The device becomes the customer

The endgame he describes is a refrigerator that opens its own ticket: it detects a developing fault, finds a five-star technician two and a half kilometres away who is free on Tuesday at 2pm, and asks the owner only to click OK. That inverts the service relationship, and it also solves distribution — an OEM white-labels lifetime maintenance into the appliance, so the consumer product arrives through a B2B contract rather than a consumer marketing budget.

▶ 9:32 ▶ 9:48 ▶ 11:40

Build in five markets at once or become a services shop

The conventional sequence is win one market, then port. Sensfix deliberately ran early proof-of-concepts in Germany, Spain, Poland, the US and South Korea simultaneously, co-developing with a handful of customers in each. The reason was defensive: if you only ever build for the customers in one market, you end up a servicing company with one client's roadmap. Building in parallel forced the feature set to stay a product — and, he says, turned the question from whether there is fit into how fast the roadmap can ship.

▶ 20:47 ▶ 21:04 ▶ 21:19

The economics live in the first visit

The metric he leads with is first-visit success rate: today a technician's first trip is largely diagnostic, after which he leaves and returns prepared. Sensfix claims that rate improves by roughly 75% when the complaint arrives pre-classified with the right skill, tool and consumable attached, that mean time to repair drops and repair failure falls below 1%. The translation for a buyer is 12–14% off annual maintenance contract spend, and as much as 30–40% in the best cases.

▶ 26:56 ▶ 27:30 ▶ 27:45

A category with no budget line is a slow sale

Because everyone experiences broken maintenance, pilots are easy to start — the pain is universal and the demo sandbox spins up in hours. The obstacle is accounting, not conviction: money that could pay for AI maintenance is already committed elsewhere, because no one budgets for a category that did not exist last year. So two to six months of POC precede the purchase order, and the bet is stickiness — once teams depend on it, they cannot imagine going back.

▶ 25:42 ▶ 25:57 ▶ 26:13

AI trained on Western facilities will not survive an Indian factory

Rather than port the stack and hope, Sensfix is finalising an MOU with IIT Kanpur explicitly to Indianise its models, alongside factory pilots in Uttar Pradesh and a pipeline of hotels, restaurants, pubs and campuses. He names the failure mode directly — it is the mistake product-minded technology CEOs make — and his read of the map is counterintuitive: Silicon Valley and Europe are his home ground but slow on adoption, while Korea, Japan, Vietnam, Singapore, Turkey and India take new technology fast.

▶ 35:50 ▶ 36:07 ▶ 36:22

Run the company on the product's own DNA

At 30–35 core people and 50–60 in total, spread from California to South Korea, delegation has become the binding constraint he admits he has not yet solved. His answer is to apply the company's own thesis internally: digitise the work, then automate it, even where today that means little more than disciplined Google Sheets. It is a candid moment — the digital-maintenance company confessing that its own operating layer is a work in progress for the next two or three quarters.

▶ 29:52 ▶ 30:26 ▶ 30:59

The numbers, drawn

What the episode measures

Every figure below was said on air — timestamps included, caveats kept.

Conversation share

portion of the hour spent on each theme

AI & machine learning24%

SaaS & enterprise18%

Founder journey15%

Sales, GTM & growth12%

Manufacturing10%

Product strategy8%

Computed from the chapter map of this episode.

What Sensfix claims it takes off maintenance spend

% of annual maintenance spend saved

As stated in conversation: savings of 12 to 14%, up to 30 to 40%, on annual maintenance contract or SLA expenditure. Company-reported customer outcomes, not independently verified.▶ 27:45

Worth keeping

Lines that stay

We are the company which starts where the IoT world ends. We pick up the data and the intelligence they have created, and we reduce your workload.

— Balaji Renukumar ▶ 32:13

The refrigerator will tell you: I am developing some issue, and I have found Kumar who lives two and a half kilometres away. He's five-star rated, he has all the things, and he comes next Tuesday at 2pm. Do you want me to get repaired? You just have to click OK.

— Balaji Renukumar ▶ 9:48

If somebody is attending to your maintenance issue with a bad mood, you will not get good service.

— Balaji Renukumar ▶ 7:42

We never built a product and then tried to push it into the market. We were on the field on day one.

— Balaji Renukumar ▶ 14:04

We don't believe the AI that we have built for the Western world would work in the Indian context — this is the mistake which many product tech-geek CEOs make.

— Balaji Renukumar ▶ 36:22

Clips that travel

Short on time? Start here

Product leaders designing AI that acts, not advises

The refrigerator that books its own repair

The clearest statement of the thesis: why stressed humans are the bottleneck, what gets digitised, and the appliance that finds its own five-star technician.

7:28 → 12:45 · 5 min ▶ Watch clip

Founders choosing between depth in one market and breadth

Field-first, and five markets at once

How a streetlight-switch hardware business became a maintenance platform, and the deliberate defence against turning into a services company.

18:18 → 22:05 · 4 min ▶ Watch clip

Enterprise sellers creating a new line item

Selling a category nobody has budgeted for

The $1.8 billion sizing, why ServiceNow and IBM Maximo do not cover this, and the two-to-six-month gap between an easy POC and real money.

22:05 → 26:56 · 5 min ▶ Watch clip

Facility, plant and service operations managers

The four numbers a facility manager buys

First-visit fix rates, mean time to repair, sub-1% repair failure and the AMC savings claim — the entire ROI argument in two minutes.

26:56 → 29:06 · 2 min ▶ Watch clip

Anyone taking a Western-built AI product to India

Indianising the AI, and the engineer wave

Why the models get retrained with IIT Kanpur, which markets actually adopt fast, and a spirited defence of India's engineering-college boom.

35:05 → 40:29 · 5 min ▶ Watch clip

Glossary

The jargon, unpacked

CMMS

Computerised maintenance management software — the incumbent category of digital maintenance tools, which Renukumar argues is aimed mostly at IT service rather than physical facilities.

AMC / SLA spend

The annual maintenance contract or service-level agreement budget a facility commits to keeping its equipment running — the line item Sensfix claims to cut by 12–40%.

First-visit success rate

The share of faults fixed on a technician's first trip rather than after a diagnostic visit and a return trip with parts — the metric the product is designed to move.

Mean time to repair

The average elapsed time from a fault being reported to it being fixed; a standard maintenance KPI he claims falls once dispatch is automated.

OCR

Optical character recognition — reading text from images, used here to digitise troubleshooting manuals, paper forms and the numbers on LCD equipment displays.

Geofencing

Drawing a virtual boundary on a site so software can register when a technician arrives and how long they spend there — one of the ground-level tricks layered on the AI stack.

B2B2C

Selling to a business that embeds your product in what it sells to consumers — here, an appliance maker shipping lifetime Sensfix ticketing inside a refrigerator.

Digital twin

A live software model of a physical asset or building, kept in sync with sensor data, so its condition can be inspected and simulated without being on site.

Connections

If this resonated, go here next

EP 170 · Deep tech

Full transcript

The whole conversation, searchable

Auto-generated captions, lightly cleaned. Click a timestamp to open that moment on YouTube.