GFI Software announced at the ZEBR.AI Roadshow in Prague that its Eloquens AI platform is now native to the KerioConnect mail and collaboration server. The integration lets more than 30,000 organizations that host their own email keep data on‑premises while gaining generative‑AI assistance for sorting, drafting and prioritising messages.
What GFI is launching
GFI’s KerioConnect, a long‑standing on‑premises mail server, will ship with Eloquens AI—a conversational‑AI engine originally delivered as a cloud service by IgniteTech. The feature is optional; administrators enable it from the KerioConnect console and can assign it to specific mailboxes or turn it off entirely. A 30‑day free trial is offered without a sales call, and the assistant works across GFI’s webmail, desktop client, and Microsoft Outlook plug‑in.
How the assistant works inside the server
When a message lands in a mailbox, Eloquens AI analyses the content within seconds. It then classifies the email into one of several buckets:
- No‑Reply Needed: newsletters, promotional blasts, and similar notifications are moved to a dedicated folder.
- Decision Required: messages that need a user response stay in the inbox and are flagged with a concise rationale.
- Draft Reply: routine queries trigger a draft response written in the user’s own writing style, ready for review.
The AI draws on a large‑language model (LLM) fine‑tuned on enterprise email patterns, and it stores all processing locally on the customer’s server, preserving data sovereignty.
Differentiators from cloud‑only assistants
Microsoft 365 Copilot and Google Gemini are tightly coupled to their respective cloud ecosystems, meaning organisations must migrate mail to those platforms to benefit from AI features. Eloquens AI, by contrast, runs on‑premises, eliminating the “move‑to‑AI” prerequisite. This approach addresses three common concerns:
- Data privacy: Email bodies never leave the organization’s firewall.
- Regulatory compliance: Industries with strict data‑location rules (finance, healthcare) can adopt AI without breaching policy.
- Cost control: Existing server licences remain in use; there is no additional subscription for a cloud AI service.
Practical benefits and potential risks
Early adopters can expect faster inbox triage, reduced manual drafting time, and more consistent reply tone. A McKinsey 2023 survey found that 61 % of enterprises were already piloting generative AI for knowledge‑work, indicating strong appetite for productivity gains.
However, on‑prem AI also brings challenges. Running large LLMs locally can strain CPU/GPU resources, especially on older hardware. GFI recommends a minimum of 8 GB RAM and a modern Xeon or AMD EPYC processor for smooth operation. Additionally, AI‑generated drafts may still require human verification to avoid accidental disclosure or tone‑mismatch.
Implications for enterprise AI infrastructure
The integration showcases a shift toward hybrid AI deployments where large language models are decoupled from public clouds. It aligns with Gartner’s 2026 forecast that “AI‑enabled edge and on‑premises workloads will grow 40 % year‑over‑year.” By embedding the model in KerioConnect, GFI taps into existing server management tools, reducing the need for separate AI‑infrastructure stacks.
From a platform perspective, the move illustrates how generative AI can become a modular service within traditional enterprise software stacks—similar to how analytics or security modules have been added to ERP systems in the past.
Business implications for MSPs and end users
Managed service providers can now bundle Eloquens AI as a value‑added service across their client base, provisioning the assistant from a single console. For end users, the technology promises a measurable reduction in email‑handling time—IDC estimates that AI‑assisted email triage can cut processing time by up to 30 %.
Nevertheless, organisations must weigh the operational overhead of maintaining AI‑ready hardware against the productivity upside. GFI’s licensing model remains per‑server, which could be more economical for firms already invested in KerioConnect hardware.
Future outlook
As generative AI matures, we can expect deeper integrations—such as AI‑driven meeting scheduling, task extraction, and cross‑application workflow automation. GFI’s roadmap hints at extending Eloquens AI to its other products, including KerioControl network security and MailEssentials spam filtering, creating a unified AI layer across communication and security stacks.
Market Landscape
The AI‑augmented email market is still dominated by cloud providers, but on‑premises solutions are gaining traction. A 2024 Forrester study reported that 38 % of midsize firms prefer to keep email in‑house for compliance reasons, yet 55 % of those firms are evaluating AI add‑ons. Competitors such as Zix and Mimecast have introduced limited AI‑based threat analysis, but none offer a full‑featured generative assistant that runs entirely on the customer’s server.
Overall AI adoption in the workplace continues to accelerate. Gartner predicts that by 2027, 30 % of all email traffic will be auto‑generated or auto‑responded by AI. This creates a competitive pressure for on‑prem vendors to match cloud capabilities without sacrificing data control.
Top Insights
- Embedding Eloquens AI in KerioConnect lets organisations keep email data on‑premises while gaining generative‑AI assistance, addressing privacy and compliance concerns.
- The assistant automatically categorises, drafts and annotates messages, potentially cutting manual email handling time by up to one‑third, according to IDC estimates.
- On‑prem AI requires adequate hardware; GFI recommends modern multi‑core CPUs and at least 8 GB RAM to run the LLM efficiently.
- MSPs can monetize the feature as a subscription add‑on, simplifying deployment through the existing KerioConnect administration console.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
