In June 2018, Praveen Kola had a fairly specific invitation for anyone interested in artificial intelligence: come to booth 612, Zone 3. RandomTrees was taking its AI solutions to SAP’s SAPPHIRE NOW conference in Orlando. The invitation had a date, a place and the unmistakable optimism of a young company with something to show. Enterprise software may lack a red carpet, but it does provide excellent booth numbering.
Kola had founded RandomTrees the previous year. He was bringing AI to an audience whose working lives already involved enterprise applications, business records and the complicated matter of making one system agree with another. Years before agents became the fashionable noun, he was looking for customers among people with existing systems and unfinished work.
That choice makes more sense when you look backward. Kola’s career runs through SAP consulting, data warehouses, architecture and analytics. He came to entrepreneurship after working on the machinery beneath a large company’s decisions. Today, based in Houston, he leads a business that applies AI to documents, industrial images and enterprise data. The through line is the work that has to happen after the interesting demonstration.
Before the agents, the architecture
Kola’s technical career began with SAP consulting at Cybertech Systems in 1997. A management role at Capgemini followed in 2000, then business warehouse work at Exxon Mobil Singapore in 2001. In 2005, he joined Hewlett-Packard as an SAP BW architect. By 2010, he was a principal architect working in data science and analytics at ExxonMobil.
The progression is revealing without requiring a heroic origin story. Consulting, architecture and analytics are different vantage points on the same organizational problem: information must become useful to somebody doing a job. A record has to mean something consistent. A report has to answer a real question. A technically elegant system still has to fit the organization paying for it.
There is a certain discipline embedded in that kind of résumé. Business software comes with inherited definitions, awkward interfaces and people who know perfectly well why an apparently unnecessary step exists. For a founder entering AI, experience in that environment offers a useful starting point. The question becomes how a new capability will fit into work already underway.
Kola’s subsequent company would bring together advisory work, implementation and reusable AI software. Seen against his earlier roles, that combination has a recognizable logic. An enterprise buyer needs help choosing a problem, connecting the technology and making the result operate. Selling the capability brings a set of obligations with it.
A founder with an enterprise address book
The 2018 conference invitation is a small but concrete marker of Kola’s transition. He had moved from holding technical roles inside established organizations to presenting a company’s portfolio to potential customers. The venue placed RandomTrees near a familiar audience. His earlier SAP experience and the conference’s enterprise focus belonged in the same conversation.
In March 2019, MOURI Tech invested in RandomTrees. Kola also took on an advisory role in data science and data engineering at MOURI Tech that year. The connection continued into public events: in 2021, he appeared alongside Jim Crenshaw in promotion for a webinar on enterprise vision AI. Another event that October concerned American Tire Distributors and Snowflake.
Those appearances supply a more useful picture of his professional world than a collection of grand pronouncements would. There are implementation partners, enterprise customers and particular technical subjects. Computer vision and data modernization ask different things of software, but both involve bringing new capabilities into an operating business. Kola’s public work stays close to that setting.
The part he wants to own
By 2025, Kola was explaining RandomTrees through its proprietary accelerators. These included Halcon AI, RunML, DocAI and AI Recipes. He placed intellectual property in the agents and assets supporting the company’s services. The arrangement gives a services business something it can develop and use across engagements while still adapting the implementation to the customer.
That is a consequential choice for a founder. Every new engagement can require close attention to a client’s systems. At the same time, repeatedly solving the same technical problem from the beginning is an expensive habit. Reusable assets offer a way to carry engineering work forward, provided they remain useful when the next customer’s process turns out to be inconveniently different.
Kola’s examples were practical. Halcon AI applied computer vision to quality control. DocAI examined documents such as contracts and statements of work. He also described TrustAI as a technical guardrail
for monitoring agents, with attention to privacy risks, bias and hallucinations. The aspiration extended beyond making an agent perform a task to watching what it did.
“Our agents are our key differentiators where our intellectual property lies in.”
Kola’s description of the agents behind RandomTrees’ services, 2025
The phrase is revealing because it locates value in something the company builds. It also creates a demanding test. A reusable component needs to be reliable enough to deserve another deployment. A founder can announce an asset once; a delivery team has to make it useful repeatedly. That tension sits inside the business Kola has chosen to run.
Six minutes, and the work around them
Consider one example from the company he leads. RandomTrees describes a logistics customer processing 2,000 invoices a day. The customer had already invested in automation, yet reconciliation and exceptions kept the work fragmented. The company reports that its Invoice 360 implementation reduced average processing time from 27 minutes to six, with payment cycles falling from 28 days to nine.
These are RandomTrees’ reported results for an unnamed customer. They describe a particular engagement. The useful detail for understanding Kola’s business is the sequence: extract the information, validate it, match it to a purchase order and post it into an enterprise system, keeping human oversight available. The value depends on the steps agreeing with one another.
Time spent processing an invoice
An invoice is a good antidote to abstraction. Reading its total is useful. Establishing that the total belongs to the right order is another responsibility. Giving an employee a clear way to resolve a mismatch is another again. The paperwork has no interest in the elegance of the model. It wants to be paid correctly.
For Kola, this is where the founder’s proposition meets the architect’s habits. The work stretches across documents, rules and existing applications. A successful implementation needs the connections between them. The numbers attract attention; the sequence explains what his company is selling. Its assignment continues after the software has recognized the words.
Teaching the system to take the next step
RandomTrees’ BOL DocVerify offering presents the same problem through shipping records. It extracts product and shipment information from bills of lading, compares those fields with energy trading and risk management records, and maps information downstream. Analysts can review and edit the results. The design includes assignment, extraction, matching and orchestration components.
That division of work matters. A shipment document can contain readable information that fails to match a master record. Someone then needs a way to see the discrepancy and act. The software’s job includes preserving that opportunity. More automation creates a greater need to make the remaining human decisions legible.
Industrial vision adds a different input. The company’s Halcon AI offering uses cameras and models for inspection in manufacturing and logistics, with reporting collected in a dashboard. In one case the system reads a document; in another it examines images. Both need a route from an observation to a person or process that can use it.
This is an editorial sketch of the recurring problem, rather than a technical diagram of a single product. It helps explain why Kola’s company combines data engineering, applications and AI. Each capability supplies a piece of the route. A customer needs the whole journey to remain intelligible.
The catalogue keeps growing
In 2025, Kola described an ambition to expand RandomTrees’ digital workforce and build more than 100 assets by that year’s end. Today, the company advertises more than 300 AI agents across industries and workflows. The figures describe different categories and moments; they are not a clean growth series. What they do show is continued emphasis on reusable software.
The organization around that ambition spans several places. RandomTrees lists its headquarters in Cypress, Texas, delivery operations in Hyderabad, an AI research presence in Chennai and a sales office in Sharjah. Its leadership includes people responsible for data, AI practice, operations and client engagement. Kola’s role as founder and CEO sits above work requiring several kinds of expertise.
Return to the invitation to booth 612. It captures an early, ordinary act of entrepreneurship: show up where the customers are and ask them to look. Kola’s company has since acquired more products, more geography and more elaborate vocabulary. The assignment remains pleasantly concrete. Somewhere in a business, information has arrived. Someone needs the next step to happen. That is where he has put his company to work.