155 Feature Updates in 2025, and What’s Coming in 2026

Enterprise AI adoption has been stuck at 10%, and the models aren’t the reason. Most employees don’t know what AI can do for them, don’t have time to find out, and the tools they’ve been given don’t help.

In 2025, we shipped a platform built to fix that. Below is what we built, grouped under the four pillars elvex is organized around: Guide, Build, Share and Control.

At the end you’ll find what we’re shipping in 2026: personal and enterprise context, power tools and code execution, task planning and sub-agent routing.

Guide: Turning Every Employee Into a Power User

The models can already do plenty. What holds adoption back is that people don’t find out what they can do. Most employees try a few prompts, get stuck, and go back to working the old way.

So this year we shipped features that start from where employees are and show them what they can do.

Guided Onboarding means new users don’t land on a blank screen. The welcome flow walks them through the main features on day one, so they get work done right away instead of waiting for a training session.

App Routing and Enterprise Search put vector search at the center of elvex. An employee with a question doesn’t need to know which assistant to use or where to look. elvex finds the right assistant, the right conversation or the right resource for them.

App Builder 2.0 walks users through building a custom agent one step at a time. Most people aren’t prompt engineers and shouldn’t need to be. You build an assistant now by describing what you need.

Self-Improvement Mode lets assistants suggest changes to themselves. As people use an assistant, it notices what’s working and proposes new rules. Every version is saved, so an editor can always roll back. The assistant points out its own improvements.

Citations now appear everywhere: in conversations, in drafts and in search results. When elvex answers, you can see where each piece of information came from. People trust the output because they can check it.

The point of all this: elvex doesn’t just hand employees AI and walk away. It shows them how to use it, what to try next and how to get better results. This is a big focus for us in 2026: read about it at the end.

Build: Removing Every Barrier to Creation

elvex is built on the idea that anyone can make AI workflows without writing code. This year we expanded a lot of what people can build.

Build on any model you want. We added more than a dozen models this year: Claude 3.7 through Opus 4.5, GPT 4.1, o3, GPT 5, 5.1, and Gemini 2.5 and 3 across the board. Different jobs call for different models. Some need heavy reasoning, some need speed and low cost, and some need a model tuned for code, analysis or creative work. On elvex, employees aren’t tied to one vendor’s roadmap. They pick the model that fits and switch when the market moves.

Create any kind of deliverable. Early in the year we launched Drafts, a workspace for documents, spreadsheets and code that sits next to your conversation. Then came image generation with DALL-E and GPT Image 1. After that, advanced charts with full ECharts support. People use elvex to produce finished work (reports, visuals, analysis and code) in one place, refining it with AI as they go, instead of only chatting.

Connect to any tool or data source. We added database connections for Snowflake, PostgreSQL, BigQuery, Redshift and SQL Server, and knowledge tools like Confluence, Notion and Coda. You can even use any website as a live data source, crawled weekly and kept in sync. Importantly, it’s both read and write access. With Actions, employees can update a Salesforce record, open a Jira ticket or send a Slack message from the same screen, without copying and pasting between ten tabs.

Automate with advanced Flows. We rebuilt Flows from scratch. You can schedule runs ahead of time, add decision steps that branch, attach files, loop over batches of data, and put flows inside other flows for longer multi-step automations. Work that used to need a developer and several weeks can now be built by anyone in an afternoon.

What this adds up to: any model, any output format, any data source, and automation as simple or as involved as you need. If your employees can picture it, they can build it.

Share: Making Adoption Organic, Not Mandated

AI pays off when something that works for one person starts working for everyone.

We see the same thing over and over. A power user builds something great, like an assistant that saves them hours a week or a workflow that takes over a miserable process. Then it stays with them. Maybe they bring it up in a meeting, maybe they don’t. Either way the value never leaves their account.

This year we made sharing easy enough that good work spreads on its own.

Groups let you organize users by team, department, project or role and share with the whole group at once. When you build a good assistant, you don’t share it with 47 people one at a time. You share it with “Sales Team” and everyone has it right away. Sharing becomes one click.

Tags make shared work easy to find. Tag an assistant “Sales,” “Competitive Intel” and “EMEA,” and anyone searching those terms finds it right away. If people can’t find something, they won’t use it.

Conversation and Assistant Sharing lets teams pick up where someone else left off. Had a good back-and-forth that cracked a tricky problem? Share it with your team, and they can copy it and keep going. Whenever you share something, the recipient gets an email about it automatically.

Mobile and Voice put elvex wherever people are working: on the commute, between meetings, on the warehouse floor. The AI your team built goes with them.

Slack is the tool employees already spend their day in. We added support for multiple Slack workspaces and shared channels, and you can now set assistants per channel, so each team gets the right one automatically.

Our view is that AI adoption shouldn’t be ordered from the top. It should spread from the people doing the work. When sharing takes no effort, the best workflows get used the most.

Control: Making IT the Enabler, Not the Blocker

None of this matters if IT can’t say yes.

Every enterprise we talk to has the same fear: “If we open this up, we lose control.” People using AI however they like, data leaking, costs running away, security holes, compliance failures. These are real reasons AI projects stall, not theoretical ones.

This year we put a lot of work into making elvex something IT can approve without worrying.

Analytics show leadership how AI is actually being used: who’s active, which assistants are earning their keep, and what people are asking about. You can see what’s working, spot what isn’t, and decide where to invest next based on data.

Custom Roles let you set exactly what each role can do, matched to your org structure and compliance rules. A new Member role lets more people build and create while sharing stays with admins. For companies on SSO, we now support SAML attribute-based provisioning, so roles and groups are assigned automatically from your identity provider.

Security and Performance work this year included patching vulnerabilities, better handling of authentication tokens, more reliable connections, and enterprise features like Snowflake key pair authentication. Audit logs got a clearer structure and more detail. elvex is SOC 2 Type 2. It is also HIPAA certified. Data is encrypted and never used to train models.

Integration Actions Controls add guardrails for when AI writes back into your systems. Flow dependency warnings stop people from breaking an automation by accident. Slack controls per channel let you decide exactly which assistants answer where. You get the power with tight control over it.

Token Counting and Cost Visibility mean no surprises on the invoice. Token counts are more accurate across providers now, and users get a warning when a conversation is long enough to affect cost or performance.

So elvex is a platform both employees and IT can like. Governance is part of every layer and was designed in from the start, not added at the end.

What’s Coming in 2026

We’re not slowing down.

Power Tools moves AI from telling you how to do something to doing it. Instead of explaining how to analyze a spreadsheet or convert a file, the AI writes and runs the code in the background and hands you the finished result in seconds. It can create Word documents, PowerPoint decks and PDFs with professional styling, clean up messy data, build dashboards, and keep adjusting them as you ask in plain English.

Context, Planning, and Sub-Agents are a major upgrade to the elvex Home App. It can now plan a complicated task as a series of steps and hand pieces of the work to sub-agents. We’re also adding context about your role, your groups, your projects and the company you work for, and through our integrations we’ll build up a picture of you and how you work quickly. The Home App knows that context, along with the assistants, flows and data shared publicly in your company workspace, so it can take you straight to the workflow that fits the problem in front of you.

You’ll be able to see how it works. We show you what we’re building and, more importantly, let you change it.

Until now, power users could do remarkable things with AI while many of their colleagues didn’t know where to begin. With these changes, everyone can work like a power user, with the AI showing them how.

elvex is the agent platform that helps every employee work like an AI native. If you’re ready to move past the 10% adoption plateau, get in touch.

Bring your agent. Bring your team.