eGain

Models carry the parts that make this current rather than legacy: conversational guidance to frontline staff, agentic orchestration, retrieval of the right answer at the right moment, and analysis of customer and employee feedback to detect emerging needs and flag gaps in the content itself.

Underneath sits a knowledge lifecycle system, the authoring, review, approval, versioning and publishing machinery that turns scattered documents into a governed single source, and that machinery is a real product the company sold long before models could read it. Strip the models and an institution still has governed content and a workflow to maintain it. This is the Perfios position, with models doing the differentiating work inside a platform whose value is also its governance.

Oversight is the product rather than a control placed around it. The answer space is bounded by governed content, with every response traceable to its approved source, so an assistant or an agent can only say what the institution has already reviewed and sanctioned, and comprehensive knowledge lifecycle management means that approval is a maintained process rather than a one time act.

The framing is stated directly and inverts the usual order, that artificial intelligence transformation starts with trusted knowledge rather than the other way around, and that the information powering the machine must be reliable and compliant first. The same foundation is explicitly designed to keep both people and systems aligned to current regulatory guidance. This is the Posh AI property of governing responses against approved procedures, except that here it constitutes the entire offering.

The strongest combination available for a knowledge product and it works on three levels at once. The answer space is bounded, since responses come from approved content rather than from model recall, which removes the fabrication surface instead of guarding against it. Every response is traceable to its approved source, so any answer given to a customer can be checked against the document that produced it and the institution can show an examiner where it came from.

And the loop closes, with feedback from customers and employees analysed to surface emerging needs and flag knowledge gaps for content refinement, so failures feed back into the corpus rather than recurring. Independent assessment corroborates it, with a leader position in the first analyst quadrant for this category and the top rating for compliance driven use. What is still absent is a published accuracy or resolution rate.

Named financial institutions across the entire size range, each with a quantified result. One of the largest global banks migrated more than 2,500 content artifacts from document, intranet and relationship platforms into the system and cut average handle time by 14 percent.

One of the largest credit unions in the country delivers proactive service to millions of members and sends over a billion messages annually through the conversation product, with interface integrations giving agents a full customer view. A 4.5 billion dollar credit union serving more than 170,000 members selected the knowledge hub in August 2026, and another is quoted on the product.

Anonymised cases add a 36 percent improvement in first contact resolution with training time cut 40 percent, and 15 percent higher advisor productivity at a top ten global bank. Independent assessment is current and specific: a leader position in the inaugural analyst quadrant for customer service knowledge management in July 2026, and the top rated solution for compliance driven service use cases. The company is publicly listed.

The architecture answers the boundary question by construction rather than by promise. The system answers from the institution's own approved content, described as a single source of truth for its policies, procedures and product information, so what the machine can say is drawn from material that institution authored and approved rather than from a corpus pooled across customers. For competing banks and credit unions sharing a vendor, that containment is the property that matters most. Held at B because nothing states whether interaction data, search behaviour or content gaps observed at one institution inform models or defaults serving another.

No data protection agreement, retention schedule or subprocessor list was located. The primary payload is comparatively benign, since the knowledge base holds the institution's own policies and product documentation rather than customer records, but the conversation product is a different matter, handling proactive outbound messaging to millions of members across text, email and mobile at a volume exceeding a billion messages a year, with interface integrations assembling a full customer view for agents. Nothing published describes how that interaction and profile data is retained or separated.

No attestation, certification, trust centre or enumerated framework was located in accessible material. Deployments at a top tier global bank and at one of the largest credit unions in the country mean security assessment has been passed at demanding standards, and listed company status brings audited internal controls over financial reporting, which is a genuine external examination of something else. None of that assurance is published for the product, so a new institution begins its own review from nothing.

Compliance is central to the positioning rather than incidental, and unusually it carries third party corroboration: the analyst quadrant that named the company a leader also identified it as the top solution specifically for compliance driven service use cases, which is external validation of a compliance claim rather than the claim alone.

The retail banking suite incorporates automated regulatory change handling from a specialist partner, and the stated benefit is ensuring people and systems adhere to the latest regulatory guidelines. What is absent is specificity, with no regulator, statute or conduct rule named anywhere, so a buyer cannot see which obligations the governed knowledge is built to evidence.

No credit or identity decision applies, so the axis adapts, and the sharper exposure is on the sales side. The retail banking suite is built to increase revenue through needs based selling, prompting frontline staff to recommend relevant and timely offers inside live customer conversations, which means a model influencing what products a customer is offered at the moment they are speaking to their bank.

Nothing describes how those recommendations are generated, whether suitability constrains them, or whether prompting varies with a customer's balances in ways that concentrate offers on the most profitable. A quieter second exposure follows from the architecture: because answers come only from approved content, a question the content does not cover is answered poorly for everyone it affects, so coverage gaps become service gaps and the platform's own gap detection is the only published safeguard.

No guarantee, indemnity or falsifiable accuracy commitment was located, and what earns the grade is that accountability is reconstructable by design. Because every response traces to an approved source, an institution can establish exactly which document produced an answer given to a customer, who approved it and when, which converts a service dispute from an argument about what was said into a documented chain, and it is the property the company sells as auditability. Gap detection means a recurring failure surfaces rather than persisting silently. What is missing is anything owed by the vendor when the retrieval itself, rather than the content, is at fault.

The composition of the banking suite is disclosed component by component, naming the specialist supplying automated regulatory change handling, which is itself indexed here, and the enterprise browser provider whose chief executive is quoted, so a buyer can see that the product is assembled from three parties rather than built by one. Source systems in the migration case are named. What is not disclosed is the model layer, with no provider identified for the conversational guidance or agentic orchestration, and no subprocessor list or hosting arrangement located.

Integration is evidenced at both ends and by migration rather than by claim. A global bank consolidated more than 2,500 artifacts out of named document management, intranet and customer relationship platforms into the system, which is harder proof than a connector list. Interface integrations at a large credit union assemble a full customer view for agents and drive automated notification workflows.

The retail banking suite composes two further named products into the platform, one supplying regulatory change handling and one an enterprise browser that removes switching between applications, with the partner's chief executive framing the requirement as institutions needing a workspace they control rather than inherit. Delivery reaches contact centres, branches, digital channels and automated assistants from the same foundation.

No hosting provider, region selection, residency commitment or private deployment option was located. The knowledge content itself is institutional documentation rather than customer records, which lowers the stakes, but the conversation product processes member messaging at very large volume and the agent workspace assembles customer views, so a regulated institution will still be expected to establish where that processing occurs.

No pricing, packaging or basis of charge is published. Listed company status means overall financial performance and revenue composition are filed and readable, which tells an observer about the business but not what a buyer pays, and the retail banking suite bundles three separately sourced components whose commercial treatment is not described. Nothing indicates whether charge falls per user, per institution, per article or on consumption.

Within banking the coverage spans the full size range, from community banks and mid sized credit unions through regional institutions to global banks, and reaches contact centres, branches, lending and back office teams rather than one function. The 2026 retail banking suite gives that its own product identity, bundling guided selling, regulatory change handling and a controlled workspace on top of the general platform.

What holds this at B is that financial services is one industry among several for a general customer service knowledge company, so this is a named banking product inside a horizontal business rather than a financial institution vendor throughout.