Instead of manually building filters, switching tabs, or jumping between tools, you can simply describe what you’re looking for in plain English and let the assistant do the heavy lifting.

“Find fast-growing SaaS companies in DACH with 100–500 employees”

“Show untouched high-fit accounts in my top segment”

“Draft outreach for acme.com using my top customers as proof points”

The assistant translates those requests into structured searches, applies your segmentation and CRM context, and returns ranked companies, contacts, insights, and outreach suggestions.

The feature is still in beta, so expect rapid improvements, occasional rough edges, and new capabilities landing regularly.

Once enabled on your account, you can access the feature from the left-hand sidebar under Chat.

When opening the chat for the first time, the center panel displays suggested workflows and example prompts to help you get started quickly.

These workflows cover use cases like:

Selecting any suggestion immediately starts the workflow.

Selecting See all prompts opens the full Prompt Library.

This is your central workspace for browsing, running, organizing, and creating reusable AI workflows.

Inside the Prompt Library, you can:

Every prompt can leverage your Ocean.io account data, including:

If your CRM or other integrations are connected, the assistant can use that data to identify ICP patterns, prioritize accounts, compare against successful customers, and personalize outreach.

At the top of the Prompt Library, you’ll find two tabs:

Curated prompts are shared across your workspace and accessible to everyone in your Ocean.io account.

These are ideal for:

Private prompts are only visible to you.

These are useful for:

In the top-right corner of the Prompt Library, you’ll find the New Prompt button.

This allows you to create reusable workflows tailored to your sales process, ICP, or prospecting strategy.

You can also organize prompts using categories and quickly locate them later using the built-in search bar.

Each prompt includes:

At the top of the left sidebar, you can start a completely new conversation or workflow.

This resets the current context and allows you to launch a fresh search from scratch.

Examples:

“Find B2B SaaS companies in Benelux with RevOps teams”

“Show recently funded fintech companies not yet in our CRM”

“Compare datadog.com to our top closed-won customers”

The assistant then converts your request into structured search logic behind the scenes.

Below the Prompt Library in the left sidebar, you’ll find the AI Context section.

Selecting this redirects you to:

Settings → Instructions

This section acts as your persistent AI configuration layer.

Here you can define:

The assistant automatically applies this context to future conversations, making searches, recommendations, scoring, and outreach significantly more relevant.

Without context configured, the assistant can still work, but responses will be more generic and require more manual prompting.

The chat is more than just a conversational UI.

Under the hood, the assistant translates natural-language requests into structured searches across:

For example, a prompt like:

“Find growth-stage SaaS companies in DACH with HubSpot and RevOps leaders”

may automatically translate into:

The assistant then ranks results based on relevance and fit.

Unlike traditional filter panels, the chat maintains conversational context.

This means you can refine searches naturally without rebuilding them from scratch.

Example:

“Find SaaS companies in DACH with 100–500 employees”

followed by:

“Only untouched accounts”

or:

“Now expand to Western Europe”

The assistant updates the existing search while preserving previous filters and context.

You can research individual companies simply by entering a domain.

Example:

“Tell me about acme.com”

The assistant can return:

Using domains generally produces more accurate lookups than company names alone.

The assistant can also search for contacts using:

Examples:

“VPs of Sales at growth-stage SaaS companies in DACH”

“Marketing leaders in the Nordics with HubSpot experience”

You can even paste a LinkedIn profile URL directly into the chat to resolve the person and associated company.

The assistant can generate outreach drafts using your account context, segmentation, and closed-won customer patterns.

This includes:

The more context you provide, the stronger the output becomes.

Example:

“Draft outreach for acme.com using our top fintech customers as proof points”

You can then refine the output conversationally:

The left sidebar also includes:

This makes it easy to revisit earlier searches, account research sessions, and outreach drafts.

Inside any active conversation, the top-right controls allow you to:

This is especially useful for internal collaboration and handoffs.

Inside the message input area, you can choose between multiple AI models depending on speed, reasoning quality, and credit usage.

Fastest and lowest credit usage.

Best for:

Stronger reasoning and more contextual understanding.

Best for:

Uses more credits than Haiku.

Most capable and highest reasoning quality.

Best for:

This option consumes the highest amount of credits.

Every prompt consumes credits.

Credits may be used for:

The credit cost of each response is displayed directly below the AI response after the request has completed.

There are currently no warnings or upfront estimates showing how many credits a prompt will consume before you run it. Because of this, broad or complex searches may use significantly more credits than expected depending on:

Previously revealed records generally do not consume reveal credits again if already unlocked earlier in the account.

Ocean.io also supports bringing your own Anthropic API key to power chat sessions.

You can configure this from:

Bottom-left profile menu → Settings → LLM Credentials

Once configured:

This setup is especially useful for:

Anthropic billing and token management remain fully on Anthropic’s side.

The right-hand side of the interface contains the live results overview panel.

This panel displays:

From here, you can:

This turns the chat into both a conversational assistant and a fully interactive prospecting workspace.

This is the single biggest improvement you can make to search quality and outreach relevance.

The more the assistant understands your ICP and positioning, the better the results become.

Detailed prompts generally outperform broad requests.

Example:

“VPs of Sales at SaaS companies in DACH with 100–500 employees”

works much better than:

“Find marketing leaders”

Domains resolve directly to Ocean entities and improve lookup accuracy.

You do not need to restart searches.

Simply continue refining naturally:

“Narrow to under 200 employees”

“Only companies not in our CRM”

“Expand to all of Western Europe”

If a filter cannot be applied cleanly, the assistant may soften or remove parts of the query.

The chat usually explains these adjustments in the response, so it’s worth reviewing them before trusting large searches blindly.

Because the feature is still in beta, feedback is extremely valuable.

For bugs, incorrect results, broken prompts, or workflow suggestions:

Email: [email protected]

You can also share problematic chat sessions directly with support to help debug specific workflows faster.