Helport AI wants to put expertise on the payroll

The expert in a contact center has a peculiar problem: everybody needs her at once. One agent has a product question. Another has a customer who has strayed from the script. A third needs the answer the expert gave last Tuesday, which nobody wrote down properly. Experience is valuable, but its owner has only one voice.

Helport AI began by putting another voice beside the agent. Its AI Assist software supplies guidance during customer conversations, while tools for supervisors, quality assurance and knowledge management help the operation around them. The premise is pleasantly practical. An employee should not need several years of institutional memory to answer a question competently.

The expert cannot take every call

Consider a mortgage inquiry. Knowing the product is only part of the job. Someone must hear the question, identify the relevant information, explain it clearly and decide when the conversation requires another person. A knowledge repository helps with one of those steps. An effective operation has to connect all of them.

That distinction explains Helport’s position in the market. It serves organizations for which conversation is a business process: selling, servicing, qualifying and collecting. The company’s expertise sits in the combination of domain knowledge and contact operations. Its older promise was to help people work like experts. Its current mission is to turn human expertise into scalable AI labor.

Agent assistance alone is a busy category. Cresta, for example, also offers live guidance, knowledge and automation to human representatives. Helport’s distinctive commercial argument is the addition of managed delivery: software, trained workflows and the people responsible for keeping the service running. A buyer is evaluating an operating arrangement as well as a user interface.

A software company acquires a shift rota

The AI+BPO service introduced in January 2025 combines AI with business process outsourcing. Rather than leaving a customer to assemble every part of delivery, Helport supplies an arrangement that includes human agents and operational support. Its regional network spans English-, Spanish-, Thai- and Bahasa-language delivery hubs.

The April 2026 QuickCEP alliance extends that approach to global brands and cross-border commerce. QuickCEP brings its customer-service SaaS platform; Helport brings AI workforce capabilities and delivery operations. The attraction is a shorter organizational journey between buying technology and having someone answer the customer. The partnership’s projected customer pipeline remains a forecast, rather than an achievement to count in advance.

Named relationships make the proposition less abstract. Helport has described support and onboarding work with Atome in the Philippines, and a mortgage-related sales collaboration with Best Life & Co. in the United States. These are different assignments with a common inconvenience: the business needs competent conversations in sufficient quantity.

The invoice tells you what changed

Helport’s change of direction is easiest to see in its pricing. AI Assist has used monthly subscriptions plus outcome commissions, typically negotiated at 7% to 15% of the gross service fees earned by agents using the system. That percentage refers to a defined service-income base, rather than a slice of every customer’s sales.

HyprX, launched in January 2026, introduced a digital-expert platform designed to reproduce professional knowledge and communication methods. Its launch pricing scaled subscriptions by the number of digital twins, interaction volume and customization. Applications include sales conversations, training simulations and preliminary intake before handing matters requiring professional judgment to a human.

Newer managed offers move the invoice closer to activity or results: qualified leads, appointments, conversions, revenue share or specified interaction volumes. The exact arrangement depends on the service. This makes the definition of an outcome part of the product. A lead that nobody can use is an expensive piece of punctuation.

“We are not selling software. We are selling industrial-grade AI labor capacity.”

Guanghai Li, CEO · April 2026

That is management’s thesis, and it changes what a buyer should ask. What counts as a completed interaction? Who verifies quality? When does the AI call for help? A plausible answer in a demonstration is one thing; an accepted, correctly recorded business event is another.

A vacuum cleaner gets a voice

HyprX for Hardware gives the strategy a particularly tangible form. Scan a QR code on a device, its packaging or its manual, and a device-specific agent opens. It is designed to answer setup and troubleshooting questions, accept photographs for visual diagnosis and pass difficult cases to a human with the conversation summarized.

The manufacturer gains a possible direct relationship with the customer, subject to consent. The customer gets a route into support without first deciphering the manual. Helport’s June launch offered outcome billing with no upfront cost to selected customers; that is a stated offer with conditions, rather than a universal free-deployment promise.

By August, the company reported live after-sales operations across North America and Europe. Named customers included Dreame and Birdfy, the smart bird-feeder camera brand. A confused vacuum owner and a disappointed birdwatcher may ask very different questions. Both benefit from answers attached to the correct product.

Hardware also exposes the limits of automation neatly. Many setup questions repeat; exceptions still need specialists. Helport reported a higher AI handling share in chat and email than in voice. The useful measurement is how many interactions are resolved without human intervention, tracked separately by channel. Counting every conversation the AI merely touches would flatter the machine and inconvenience the accountant.

Revenue grew. The cushion shrank.

Helport’s fiscal 2025 revenue reached approximately $34.9 million, up 17.9%. But cost of revenue grew faster, and net income fell from about $7.4 million to $1.9 million. The company attributed higher costs partly to outsourced operations and software amortization associated with new markets and applications. Expansion had a price.

The six months ended December 2025 brought $17.7 million in revenue and a $1.7 million net loss. Research and development spending rose, and receivable credit losses also weighed on the result. These figures complicate the notion that an AI business automatically escapes the costs of people, deployment and collection.

Financing has supplied its own awkward lesson. At the August 2024 Tristar combination, a planned $15 million private placement delivered $5.5 million because one investor could not remit substantially all its subscription. Convertible notes brought combined gross proceeds to $10.39 million. The commitment and the money in the bank were very different numbers.

In July 2026, Helport established an at-the-market facility for up to $9.55 million. That is selling capacity, not proof the amount was raised. It also disclosed a Nasdaq minimum-bid-price notice, with a compliance period through January 11, 2027 and no immediate effect on listing. The operating thesis and the public-market reality deserve separate scorecards.

September brings a sharper test

A white paper published on October 1 supplies a more specific picture of the work. In a Philippines collections program during September, Helport reported revenue growth of 15.71% and an approximately halved training-and-launch cycle. In Mexico, it reported 30% more cases handled per worker and 40% more revenue per worker during the project period.

Those are company-reported deployment observations. The company says they were not independently verified in a controlled study and may reflect case mix, policies or market conditions. They describe those programs, rather than a dependable percentage for the next customer.

The underlying method is more useful to borrow than the percentages: study successful interactions, test approaches in simulation, execute across business systems and feed production outcomes back into training. The experiment continues after launch. A business process has customers, records and consequences; keeping all three in view is rather more demanding than producing a fluent reply.

Copy the workflow, then count the exceptions

For a reader considering this approach, the sensible starting point is a repeatable job with an observable finish. Choose a bounded support question or qualification step. Record the expert’s method. Specify when to escalate. Measure correct completion alongside cost, repeat contact and human intervention. These are practical deductions from Helport’s operating model, not promised customer results.

The fit becomes weaker when the knowledge is unsettled, the cases rarely repeat or the outcome cannot be agreed. Work requiring professional judgment also needs a clear human route. Helport’s own hybrid architecture makes that boundary visible. The central question is whether expertise can be reproduced reliably enough to carry an operation - including the customer who refuses to follow the script.