How AI Is Transforming Payroll & Workforce Management in Construction - Shreesha Ramdas (Lumber CEO)

Rishi Srivastava (00:43)
Today our guest is Shreesha Ramdas. Shreesha, welcome.

Shreesha Ramdas (00:46)
Thank you, Rishi. Pleasure to be here.

Rishi Srivastava (00:48)
The theme of today’s episode is AI, Workforce Management, and the Future of Construction Finance. The first section here is on origin story and vision. Tell me a little bit about yourself. What’s been your journey leading up to founding Lumber?

Shreesha Ramdas (01:01)
Yeah, so I’ve been building SaaS startups, Rishi. I built StrikeDeck that was acquired by Medalia. And before that, I built Lead Formix that was marketing automation that was acquired by SAP Caledas. And so a lot of my focus last few years has been on workflow automation, especially related to customer journey.

So with Lead Formix, I was focused on lead to customer journey. How do you get leads and how it progresses to becoming a customer. With Strike Deck, my next startup, I focused on how do you take a customer and convert them into successful customer. So startup is definitely my thing. I’m really passionate about building startups. It’s an adrenaline rush that you feel when you build something out of nothing.

there did not exist anything and suddenly it’s providing producing value for organizations. And then after I came out of Medalia, was Rishi ⁓ thinking that I’ve served enough in terms of to the technology vertical and I wanted to go into a different industry. And I wanted to go into an industry that has still not witnessed the true power of digital transformation.

Rishi Srivastava (01:50)
Mm-hmm.

Shreesha Ramdas (02:11)
And I stumbled upon in some ways construction industry. And I felt here’s the construction back office that was still struck on spreadsheets and manual work, especially payroll compliance and credentialing. And so that’s how I created Lumber where we combine these domain problems with purpose-built AI so that contractors can be accurate auditory and essentially make their back office competitive strength.

That was how Lumber came about.

Rishi Srivastava (02:38)
You raise such a good point. There’s only one industry that is behind in digital transformation compared to construction. That’s agriculture. So we are really far behind. We are living in the early 2000s here in construction. The next question here is, you built and scaled multiple SaaS companies before Lumber. What gap did you see in construction that made you want to tackle workforce management?

Shreesha Ramdas (02:47)
Yes.

Yeah, as I said Rishi, when I first started to talk to contractors, I thought, let me solve the payment problem, which is the general contractor to subcontractor or the subcontractor to material suppliers. They have to make payouts and let me solve that problem. But as I spoke to construction firms, I felt like every CFO controller, they have duct taped that problem and it’s working somehow. But

Again and again in the conversation, the topic that came up was how difficult it is to recruit workers and how difficult it is to maintain or retain skilled craft workers. And that’s when, you know, I would say it was that light bulb movement where I said like the only way to make an impact in this industry to move the industry forward.

is how do you enable construction firms to get enough workers to do and execute the projects and then retain these craft workers so that there is predictability and continuity from projects to projects and that was inspirational and definitely as we were talking to construction firms they all mentioned about how truly this industry you

have hyper fragmented data like field time, credentials, contracts, they’re all data spread in different systems. And then this industry is regulated, super regulated that there is high compliance risks. And the current systems are not built for that. So that’s where I saw an opportunity to truly unify these workflows, remove the friction that causes cost overruns and compliance failures. That’s how it all came about.

Rishi Srivastava (04:38)
How did your background in customer success and automation shape your approach to solving complex workflows like payroll and compliance for contractors?

Shreesha Ramdas (04:48)
Yeah, so Rishi customer success taught me a big lesson on change management. How difficult it is change management for organizations that are adopting new processes, new systems, new technologies. And you know, if you peel the layer a little bit, you’ll understand why change management is a big thing is because we humans, are creatures of habits.

Rishi Srivastava (04:54)
Mmm.

Mm-hmm.

Shreesha Ramdas (05:12)
We develop habits and that becomes our comfort zone and it’s very difficult to come out of that comfort zone. We are used to a workflow every day and every day, every week, every month. And to change that, it’ll require a lot of effort, a lot of determination, a lot of discipline. So I would say that actually in some ways customer taught me to design products around human workflows and failure modes.

And then automation in general, and I believe like both my last two startups were around automations and as I said, automating the customer journey. So automation in general told me how to encapsulate rules and edge cases, how to propel a motion forward. So those experiences actually helped me

Rishi Srivastava (05:55)
Mm-hmm.

Shreesha Ramdas (06:02)
build lumber to be empathetic to the operations and rigorous in controls. So that’s the reason at lumber we take audit trails, anomaly detection and integration with other system of records very seriously.

Rishi Srivastava (06:15)
Yeah. You know, you talked about edge cases If you’re not solving for them and not careful, that’s no good.

Shreesha Ramdas (06:25)
Absolutely.

Rishi Srivastava (06:26)
So the next section is building Lumber. Lumber unifies payroll, time tracking, and onboarding. All massive main points for contractors. Which workflow was the hardest to automate, and what did you learn from it?

Shreesha Ramdas (06:40)
Yeah. So I would say of all the things that we have developed and we have time tracking, have payroll, we have benefits, we have onboarding, applicant tracking system, pay card now, expense management. But I would say of all the modules, the workflows that I talked about Rishi, prevailing wages and certified payroll, they are all together, we’re the toughest. They were toughest because they are full of exceptions.

There are local rules that need to be adhered to and audit expectations. So the learning for us at Lumber dealing with this complexity was that accuracy is not just about logic. It’s also about traceability. It’s also about operators putting trust. so building really transparent review flows, anomaly detection, making it easy exports for audit purposes.

Rishi Srivastava (07:20)
yeah.

Shreesha Ramdas (07:30)
has been a major, major learning and we have been like obsessed, focused on how do we enable this in our product, in our platform.

Rishi Srivastava (07:37)
Yeah, you know, to think certified pedal, definitely government is involved there. So you definitely got to be careful as a software provider. So go ahead.

Shreesha Ramdas (07:46)
Yeah, absolutely Rishi and in

fact, you know, I will, I will, you know, elaborate, right? Like to, to showcase how difficult it is or how complex this is, right? Like if you have to reconcile contractors have to reconcile time sheets with their fringe benefit allocation and a small classification difference will trigger large back wage risks. So, so that’s that fear that contractors are in like what

Rishi Srivastava (08:00)
Cool.

Shreesha Ramdas (08:10)
What have they done in past that could trigger back wage risk? And it’s also about workforce experience, If workers are not satisfied, they are dissatisfied in how they are being paid out, they’ll not show up for workers and that again, you know, not having enough workers to do the work that they have taken up is a major problem.

Rishi Srivastava (08:30)
Yeah, so many people are retiring out of this industry. There’s still a blue collar worker gap here.

You mentioned Lumber runs on a purpose-built LLM. Can you break down what that actually means for users, especially around compliance and payroll automation?

Shreesha Ramdas (08:45)
Yeah, so what we have done here is we have trained language and reasoning layers specifically for construction, payroll, compliance and workflows. So what that essentially means Rishi is that instead of a generic assistant, this LLM understands job classification, it understands union rules, it understands payroll anomalies and we can surface explainable action. And for example, flagging a timesheet

that violate state over time rules. And then not only flagging that, also bringing the rule that’s needed, that’s applicable rule in this scenario, and then the required fix. All of that can happen together. So purpose built means we have domain data encapsulated, we have rules and auditability, and it’s not just a black box.

Rishi Srivastava (09:28)
Can you throw some more light on the training ⁓ process of this purpose built LLM? What sort of data you need and how you ensuring that high accuracy and traceability is there?

Shreesha Ramdas (09:41)
Yeah, so let’s take time tracking, right? All the teams, time sheets that workers file, right? Rishi, it’s mind boggling how valuable and useful data is right there, right? Because the workers are not only clocking time, they’re also inputting any safety incident that happened during the job. There are also data that accounts for productivity, right? So…

time tracking data, the field data, field notes, job codes, pre-filled certified payroll reports, all of that come together to ensure that the domain data is there for the LLM to process and to get better.

Rishi Srivastava (10:21)
Yeah, Yeah, so the domain data training is so important, right? Because I was talking to another speaker, who’s a construction controller just before you. And he was talking about the nightmare he’s having with payroll, right? He thinks that the payroll cannot be automated with AI. And you are like building something which is non-consensus here. And hopefully you’re right.

Shreesha Ramdas (10:42)
No, is, you know, we have had instances Rishi where we have demoted our platform to the controllers and CFOs. And some of them said this is black magic, right? And because, you know, if you think about like the, how much they need explainability, how much they need audit trails, right? Because they need to make traceable decisions. That is where, you know, you can start in a small way and then you can take on more and more.

complex scenarios that can be, where you can leverage an AI first platform.

Rishi Srivastava (11:12)
Definitely. Most construction ERPs are decades old. We’re talking about 20, 30, 40 years old. Those are the architectures. How do you think about integrating modern AI into these legacy systems without overwhelming finance teams?

Shreesha Ramdas (11:26)
Yeah, so Rishi, the right way to think about our ERP integration is that we at Lumber, we treat ERPs as a system of record and Lumber as a system of action. So we read, we write where possible. More importantly, we continuously reconcile rather than replace ERP data. So what that does is that minimizes change for the finance teams.

Rishi Srivastava (11:37)
Mm.

Shreesha Ramdas (11:51)
While at the same time it helps us unlock AI driven alerts, pre validation and reconcile job cost flows. So we essentially help the finance teams, the controllers, the CFOs maintain one source of truth. So ERP is the authority for ledgers and Lumber automates that ledger with payroll and time intelligence. Right? So that’s where we help

the finance team build trust with their own data. That’s how we look at it.

Rishi Srivastava (12:21)
I really like the system of action idea you just threw away. Can you tell me more about how you think about a system of action?

Shreesha Ramdas (12:29)
Yeah, you know, our emphasis has been that how do we, you know, rather than being the system of record, we be the system of action where the finance team is actually not consuming just reports, but they are acting on the alerts, the anomalies that is detected. So that is where in lumber, if, as I mentioned, if a worker has keyed in data,

which is the time inputs that the worker has provided. If it’s in violation of some rules, it’s detected immediately and immediately an action can happen. Rather than waiting for the payroll to happen to detect that anomaly and then going back to the worker, then having that detailed discussion with the worker, then getting the sign-offs and then implementing it. There’s a long delay. So we are like…

Can the action happen immediately as soon as an activity has happened? So that’s the perspective that we bring to the table.

Rishi Srivastava (13:24)
This is such a great point. You want to catch the problem super early on, So that there’s not a bunch of downstream consequences and so much labor wastage on that issue. ⁓ The next section here, Shreesha is on construction finance evolution. From your perspective, how has the role of the construction controller or CFO evolved in the past five years?

Shreesha Ramdas (13:35)
Absolutely.

Yeah, traditionally I would say the controllers have been historical reporters. They just report, used to report. But now I view their role as strategic risk managers and they need real-time visibility into the labor cost drivers. They need visibility into compliance exposure and they need visibility into project profitability. So now with help of technology, they can truly be

orchestrators of process and not just be number checkers. So now they, you know, given how construction industry is moving, this demand for daily weekly labor variance reporting, right, before controllers are being tasked to quantify compliance risk, right as financial exposure. So that’s the reason why

Rishi Srivastava (14:18)
Hello, sir.

Mmm.

Shreesha Ramdas (14:37)
We enable and there’s no other choice for CFOs and controllers to be now treat themselves as risk managers and not just the reporters.

Rishi Srivastava (14:46)
Yeah, They have to look through the windchill not just with the review. Many CFOs struggle with data fragmentation, payroll in one system, workforce management, and other job costing in a spreadsheet. How does Lumber bring that data together for real-time decisions?

Shreesha Ramdas (14:54)
Absolutely.

Yeah, and we covered a little bit of this Rishi in our previous conversation. So lumber in just time, as you know, we do time tracking. We have ⁓ credential management where all the experience and training certificates of the worker is there in our system and field productivity data. We reconcile it with ERP job codes and then provide reconciled views for the real time decisions so that

you know, on the project on the site can be done, right? So the CFOs actually see the labor costs by projects today and they don’t have to wait for the payroll process to be done next month, right? So that’s where the integration with the ERP with our time tracking data, the mapping layer we provide for job codes and cost buckets, the continuous validation that we do with the timesheet reviewer and anomaly detection and our pre…

validated payroll run, certified payroll report, all of that brings all of that data together for the CFO and the controller so that they can then make decisions in real time.

Rishi Srivastava (16:08)
Contractors operate on razor thin margins. Can you share an example of a financial insight or efficiency gain a customer uncovered by unifying workforce and financial data?

Shreesha Ramdas (16:21)
Yeah, one example Rishi is one of the customer that we have. They found three to 5 % labor leakage reduction after we automated time classification and caught actually a misallocated over time and miscoded billable hours. And that flowed straight to improve gross margins on the project. So that’s an example of how we bring it all together.

Rishi Srivastava (16:30)
Soon.

Mm-hmm.

Shreesha Ramdas (16:48)
for the controllers, right? If the field crew locks over time, but that was not coded to the right job, then we flag that pattern and correct the locations, right? And so that then translates into immediate job profitability.

Rishi Srivastava (17:05)
Yeah, think the payroll is such

a manual process right now, at least the people who are not your customer. So, know, this kind of time saving and error reduction that you’re talking about is almost a magic. The next section here is on leadership and culture. You’ve led and scaled teams at multiple startups. What’s your philosophy for building a culture of

innovation inside a construction focused company.

Shreesha Ramdas (17:30)
Yeah, so I can go on on this topic for hours Rishi. So one thing, know, having done few startups, have enough scars on how to hire, recruit star performers, how to build team and culture is truly close to my heart. I’m a big fan of Ben Horowitz book on the subject called culture, which is like culture should be implemented from day one of the company, right?

Rishi Srivastava (17:34)
Mm-hmm.

No, no, no.

Shreesha Ramdas (17:54)
If as a CEO, if I’m coming late to the meeting, be it five minutes, 10 minutes, then that becomes a culture, right? The management team will then follow and then the employees will follow, right? So at Lumber specifically Rishi, our big theme about culture is we have to respect the craft and the people who do it, which is the craft workers. That’s our North Star. In fact, we have posters of craft workers, right? Wherever we work.

Even our remote employees have said, please print out this poster and look at them every day because this is the segment that we are serving, right? We building products, we are serving their cause, we are empowering them. Our job is to uplift their daily life, right? So apart from that, we also try to hire operators who know the pain in the field on a job site, right? They have felt it and then pair up

pair them up with our product and engineering. So that enables us to make product calls, product decisions that are anchored in real problem and we are not building hypothesis in an ivory tower. That is one part of it. The other thing that we have been doing and I’m really proud about this is we rotate our product team, our product team members into customer success roles.

so that they get to talk to customers on a frequent basis. It’s a mandatory thing. It’s not an option. So that is one. And then we celebrate wins with our customers. So any payroll run, it’s an occasion for joy because we have done. Anytime they are audited and we are able to deliver them success in that audit, it’s time to celebrate. Celebrate together as one team, not as a customer and a vendor. And then…

I would say at the foundational level, we make all of our product decisions based on evidences. Like what are the customer metrics we are able to move? What are the gaps in the customer metrics that customer is pointing us to? What are the real life case studies that we have to enable? So that’s how we do. We have built our culture.

Rishi Srivastava (19:57)
Yeah, like having customer success people as product people and the product people as customer success people. Wow, I like that idea.

Shreesha Ramdas (20:06)
Absolutely.

Rishi Srivastava (20:06)
How do you balance speed of innovation with the reliability expectations of finance leaders who depend on precision?

Shreesha Ramdas (20:14)
Yeah, this is really an important question Rishi, especially given that we are in the payroll business. So we move fast, but I’ve absolutely emphasized that we move fast, but with guardrails. Every automation that we ship, it has to have human in the loop review. It has to have explainability. It has to have an audit log and it needs to have a rollback plan.

Rishi Srivastava (20:20)
Mm-hmm.

Shreesha Ramdas (20:38)
For finance leaders, we know that reliability is actually non-negotiable. Innovation is important, but it must be incremental and reversible if required. So we have staging environment, we have releases for critical automation, we have automated recommendation, and the human approval go hand in hand. So that’s how we balance the speed with reliability.

Rishi Srivastava (20:44)
Mm-hmm.

So the audit on your LLM models, how did you design them and thought about them?

Shreesha Ramdas (21:10)
Yeah, the audit layer, again, it’s about what are the workflows that we are automating and what sort of diligence those automations require. It’s based on that. So some of the workflows may not require that much diligence. Some of the decisions need that elaborate explainability so that we understand exactly how it is happening.

And I can give you an example. for example, we have an agent that converts manual time sheets into digital record. And I’ll tell you that some of the handwriting that we have seen at field is crazy. Right? There would be box for entries and the writing would be outside the box. And so we had to really work hard in auditing the initial result and essentially use a human in loop to

provide inputs on to correct it. And so we started with 83 % accuracy and now we are at about 95 % accuracy with so many iterations and looking at so many manual entry data. So that’s an example of how we have done.

Rishi Srivastava (22:13)
Thank you. The next section here, SHreesha is on technology and AI. There’s a lot of hype around AI in construction. Where do you see real near-term impact versus just buzzwords?

Shreesha Ramdas (22:25)
Yeah, so it’s a good question, right? So right now, truly, we are living in the age of AI. There is a lot of hype on what AI can do. And at the ground level, I have been hearing reports, not only from construction, but from other industry, that they are not seeing the ROI. Right? It’s not moving the needle for them. For us, we first listed ⁓

the short term real impact AI problems that we should go after. So automating compliance check, which is over time, the different state rules, the prevailing wage was where we first deployed the AI to work hard, Certified payroll generation, audit readiness, that was again another area. Intelligent timesheet reviews and anomaly detection, we talked about from manual time.

entries, how do we convert it to digital record? There are workers who have been on the field for several years who are resistant to change. They are not ready to have a time tracking app on their mobile app. They still prefer writing. So how do we apply AI there? It was an important thing, right? Because again, going back to change management, you have to make change management easy for construction companies for them to adopt technology. So there are younger workers

who like to be on their mobile phone, who like to consume applications on their mobile phone. So we give them a mobile time tracking. Whereas there are some of the workers who are used to doing manual time sheets, they’re refusing to change. So give them their comfort way, which is take those manual time sheets and convert them into digital records using an AI agent. So

where I would call out hype is ⁓ you cannot replace domain expertise. It’s augmentation, not magic. That’s really important to understand. Like 100 % autonomous payroll with operator review, maybe in one day, we may have automated payroll admins, but today it’s unrealistic for regulated payroll.

because the regulation is so intense and so complex. So that is where I would like to call out hype and we have to really think about the impact on the field, impact for the back office, how you can enable back office to be competitive differentiator for construction companies.

Rishi Srivastava (24:41)
Yeah, I really like the augmentation versus trying to replace people. know, like computers were a big augmenter, right? The same way here, AI is supposed to be augmenting, you know.

Payroll and compliance involve endless complexity, multi-state rules, prevailing wage, overtime. How close are we to truly automated compliance through AI?

Shreesha Ramdas (25:06)
Yeah, as I referred Rishi, I think we are close to highly automated compliance. would say about 80 to 95 percent even routine checks can be automated. But full end to end autonomy, I would say still requires human oversight because of how complex CBS can be. Imagine a 300 page

collective bargaining agreement with a clause buried in there somewhere which says if my worker works with an iron worker, they need to be paid 1.5 times their normal bonus. So those kind of language nuances, union negotiation, one of rulings, they require human oversight. Where AI helps is rule application, pattern detection.

Prefills audit artifacts. That’s where AI will help right where human will still lead is legal interpretations contract exceptions. Those are where your human judgment is needed.

Rishi Srivastava (26:10)
What’s the biggest misconception construction companies have about AI automation right now?

Shreesha Ramdas (26:16)
Yeah, this is an industry wide phenomena, fear, you know, at some levels and outside from the technology world that AI will replace operator judgment. That’s the conception. In fact, opposite is true. AI will multiply the value of experienced operators by removing repititive work.

highlighting exceptions where human judgment really matters. So that’s the perception change I think is needed by the industry.

Rishi Srivastava (26:45)
Yeah, you know,

it’s just a tool, right? At the end of the day, it’s supposed to be helping people get more efficient, do things faster, right? Or more accurately. And we think about all this progress in last 30, 40 years that we have, know, technology has been a big multiplier. know, like people were so afraid of computers in 90s. You remember? Computers are going to come and take all the jobs. And what are we going to do if the computers do all the jobs?

Shreesha Ramdas (27:11)
Absolutely, absolutely. I remember Rishi back in the dot com and I know I’m dating myself here. People were reluctant to have their bank information or access bank information on the internet. And today it’s we don’t step into bank for months together. It’s all transacted virtually, right?

Rishi Srivastava (27:19)
Mm-hmm.

Yeah.

Yeah.

All right. The last section and the last question here is on future outlook. Looking ahead five years, what does a connected back office look like for a construction company? And what advice would you give to CFOs or controllers starting that journey today?

Shreesha Ramdas (27:52)
Yeah, so I’ll talk about the vision first Rishi. I feel a connected back office is real time. Real time is important here Rishi. Real time labor costing, job costing, automated compliance, credentialed crew on every job. Every job has workers you know are credentialed, are authorized to operate the crane that they supposed to operate. Or they have the training.

to do the electrical job that they have been tasked with, right? And it’s not dated. It’s that. And one that reconciles that data with the ERP, which is the system of record, right? And that enables daily margin visibility and predictive labor spent. And that in turn, feel Rishi would move the construction firm and industry forward by reducing surprises. And it will turn labor into a managed

properly managed asset. And in terms of advice, would say three advice I would have for construction firm owners who are now embracing technology AI is start small. Pick one high value workflow like your certified payroll and automate it end to end. Protect your ledger is my second advice. Your ERP

is a single source of truth. Don’t try to mess with that, right? And let it generate pre-validated journal entries. And the last, the third is measure outcomes. Track, you know, how fast you can close your payroll, how fast you can do your audit exceptions, how far you can predict your margin changes, right? You then show your CEO how you can do the ROI.

Right? How you can generate ROI. And so I often say this Rishi, I’ve said this to my team, I’ve said this to my customers. We don’t sell AI. We sell predictability and auditability powered by AI. Right? There’s a distinction. You just don’t sell AI just for the hype, right? Because it has everybody’s attention. We are dealing with back office. are, we feel the pain that CFO and controller feel.

They will be able to sleep better when their payroll is predictable. So we should make the back office auditable, explainable and more importantly faster Rishi.

Rishi Srivastava (30:05)
Yeah, I really like the idea on the focusing on predictability. know, AI inherently is an uncertain type of technology, So the stuff that’s coming out of it may or may not be right. We need someone to be able to review it.

Shreesha Ramdas (30:21)
Absolutely.

Rishi Srivastava (30:21)
It was great chatting with you. Thank you so much for your time.

Shreesha Ramdas (30:25)
was a pleasure Rishi. Thank you.