In short: In 2026 a first sales email gets read when it shows you looked into the person and their company. The hybridAI Outreach Tool does that research in under a minute: it finds the email address, researches the person and the company with sources, lists concrete hooks for your offer and writes the email from it. Five requests are free.
Step
By hand
With the Outreach Tool
Find the email address
guess, try several tools
with a confidence score and the LinkedIn profile
Research person and company
about 20 minutes
with sources, in parallel
Hooks for your offer
in your head
listed concretely
Write the email
yourself
subject, personal greeting, reference, next step
Total time
20+ minutes
just under 1 minute (our test)
Cost
your working time
5 requests free
Anyone in sales knows the difference: an email that opens with “Dear Sir or Madam, we are a leading provider…” gets deleted. An email that shows someone actually looked into the person and their company gets read. The difference is research – and research takes time: who is this person, what is the company doing right now, where does my offer fit, and what is the email address in the first place?
The Outreach Tool by hybridAI does these steps together.
Who do you want to reach – and with what?
You only enter a few things: the person’s name and company, optionally the domain, your offer, your name as the sender, a template – cold outreach, catch-up, re-activation or warm intro – and the language of the email. In the video, the tool researches Stephan Noller, founder of hybridAI, with his consent.
Then three things run at once: the search for the email address, the research on the person and the company, and the writing of the email. In our test, it took just under a minute.
Address, research, hooks
First the address – with a confidence score and the LinkedIn profile. Below that, the research: what is publicly known about the person and what the company does, each with sources. That is more than a business card; it is what a good first conversation is made of.
The most important part are the hooks: concrete points where your offer fits the person’s situation. Exactly the work you would otherwise do in your head after twenty minutes of research.
The email – and what comes next
From all of that, the email is written: subject line, personal greeting, a reference to what the research turned up, and a clear next step. You can copy it or open it straight in your mail client or Gmail. As with any generated text: read it, adjust it, then send it.
For working with more than one person, the tool also comes as an LLM skill. The package teaches an AI agent to use the hybridAI Outreach API – then your own agent researches and writes for a whole list at once. If you sorted your network with the LinkedIn Network Radar first, you already have that list.
Frequently asked questions
What does the Outreach Tool cost?
Five requests are free, no sign-up: hybridai.one/tools/outreach.
What kinds of emails does it write?
Four templates: cold outreach, catch-up, re-activation and warm intro – in the language you choose.
Can I use it for a whole list?
Yes, as an LLM skill: it teaches your AI agent the hybridAI Outreach API, and the agent then researches and writes for every contact on a list.
Can I send the email unchecked?
No. As with any generated text: read it, adjust it, then send it.
In short: Your best lead list is often your own LinkedIn network – LinkedIn just does not show you who in it fits your offer. The hybridAI LinkedIn Network Radar reads your LinkedIn data export entirely in the browser, scores every contact for your offer (relevance 1 to 10), sorts your network into outreach segments and writes the first message. It is free, and your contact data never leaves your browser.
Segment
Who is in it
What you do with it
Hot leads
high score and already in touch
follow up now
Decision makers
C-level and VP level with high relevance
reach out deliberately
Dormant contacts
relevant, but never messaged
re-activation
Warm intros
contacts with a recommendation or endorsement
ask for an introduction
Most of us have a few hundred, often a few thousand LinkedIn connections. Former colleagues, past customers, people we met at a conference – and a surprising number who today make decisions exactly where our offer fits. The problem is not that these contacts are missing. The problem is that we cannot see them.
LinkedIn itself does not help much. You can scroll through your connections, but you cannot sort them by who is relevant for a specific offer, who you have already talked to, or who has gone quiet for two years. That is what we built the LinkedIn Network Radar for.
Your export as the starting point
LinkedIn lets everyone download a copy of their own data (Settings → Data privacy → “Get a copy of your data”). You drop that ZIP file into the radar. The analysis runs entirely in your browser – your contact data is not uploaded to a server for it. If you just want to look around first, you can start with the tool’s demo data, without an export and without signing up.
Before the dashboard is built, you tell the radar what it is about: what you offer, and which industries and positions you want to reach. Everything else follows from that – the score of every contact as well as the messages.
From a network to segments
The dashboard first shows where you stand: how many contacts you have, how many you have exchanged messages with, and how many count as high-value targets. Below that, your network is broken down by industry and position.
The outreach segments are where it gets interesting. The radar sorts your contacts into groups you can work with right away:
Hot leads – highly scored and already in touch,
Decision makers – C-level and VP level with high relevance,
Dormant contacts – relevant, but never messaged,
Warm intros – contacts with a recommendation or endorsement.
One click on a segment filters the contact list. Every contact has a relevance score from 1 to 10, a relationship strength and the right channel.
The first message is already written
The part that saves the most time day to day: click a contact and the message is ready – with the right template, addressed personally, in English or German. There are templates for re-activation, cold intro, follow-up and warm intro; adjust them once and they apply to every contact. You can copy the message or open it straight in your mail client.
You collect contacts in your own lists, such as “Decision makers Q4”, and export them as CSV or as an email list.
Your network as a skill for your AI assistant
One button is easy to miss, but it may be the most important one: Create LLM skill. It packs your analysed network into a ZIP – with a SKILL.md, the enriched contacts, a search script and a tool definition. With it, an AI assistant knows your LinkedIn network and can answer questions like “Which C-level contacts do I have in finance?”.
That way you can – if you want to – bring your network into our own assistants too: into HybridClaw, Hy or the Sales Companion. With your contacts there, you can ask the assistant before a meeting who you know at a company, or have it put together a list for your next campaign.
For quick questions, the radar also has a network assistant: “Which decision makers do I have at banks?” – and the answer comes from your own contacts.
Frequently asked questions
Are my LinkedIn contacts uploaded?
No. The analysis runs entirely in the browser; the contact data is not uploaded to a server.
How do I get my LinkedIn export?
In LinkedIn under Settings → Data privacy → “Get a copy of your data”. Drop the ZIP file into the Radar. To try it, there is demo data – no export, no sign-up.
What does the Radar cost?
Nothing – it is a free hybridAI tool.
Can my AI assistant work with the network?
Yes. “Create LLM skill” packs the analysed network into a ZIP with a SKILL.md, the contacts, a search script and a tool definition – for HybridClaw, Hy, the Sales Companion or other assistants.
Try it
The LinkedIn Network Radar is part of the HybridAI Enterprise Assistant, and you can try it with demo data right in your browser: hybridai.one/linkedin-radar.
In short: ChatGPT, Claude and Gemini send their own web searches before they answer – and those often differ from what the model says it would search for. Those real searches are what you need to be found for in 2026. The hybridAI GEO Sniffer shows both side by side, plus the websites the answer relies on, the brands it recommends and a GEO strategy. Up to 15 requests a day are free.
What the GEO Sniffer shows
What it is
What you need it for
Asked
what the model says it would search for
how the model understands the question
Live
the web searches it actually sends
the queries you need to be found for
Sources
websites the answer relies on
the places you need to be present
Brands
products and providers it recommends
your competition in AI answers
GEO strategy
keywords, platforms, missing content, next steps
the to-do list for next week
What makes a page citable was examined in an analysis of 193,507 AI answers from ChatGPT, Google AI Mode and AI Overviews (August–October 2026) (AI Citation Study). It measured associations, not causes:
Trait
Citation odds
Year in the title
more than twice the odds
Updated in the last 12 months
+88 %
Tables
+69 % (+88 % in ChatGPT)
First-hand experience
+45 %
Prices on the page
+28 %
FAQ section
+26 %
Direct answer at the top
20 % of cited vs. 5 % of passed-over pages
When people used to look for a tool, a service provider or a piece of software, they googled it. Today more and more of them ask ChatGPT, Claude or Gemini instead – and get not a list of ten links but an answer with three or four recommendations. If you are not named there, you do not exist for that question.
For businesses, that is new territory. With Google, you roughly knew how it worked: keywords, rankings, backlinks. With AI assistants it is less transparent. They often search the web themselves before they answer – but what exactly they search for, and which pages they rely on in the end, you cannot see. Generative Engine Optimization, or GEO, therefore starts with a simple question: what actually happens between the question and the answer?
Ask the way your customers would
That is where the GEO Sniffer comes in. You pick a model – ChatGPT, Claude or Gemini – and ask a question the way your customers would. In our video, it is: “Which CRM software works best for small agencies?”
The result comes in two columns, and the difference between them is what is really interesting:
Asked – what the model says it would search for,
Live – the web searches it actually sends.
The two lists are often similar, but rarely the same. And it is the real searches you need to be found for.
Who the AI recommends – and what it relies on
Below that are the websites the model relies on for its answer and the brands and products it recommends. In our example, those were comparison and review sites on one side and the well-known CRM vendors on the other. The results are live and change a little with every request – just like the answers real users get.
That leads to a very concrete insight: these sources are the places where you need to be present if you want the AI to name you.
From analysis to a GEO strategy
It does not stop at the analysis. From the queries, sources and brands, the GEO Sniffer builds a GEO strategy: the keywords that matter, the platforms where it is easiest to get listed, ideas for content that is still missing, and concrete actions for the coming week.
And because every model searches differently, comparing is worth it: the same question to Claude or Gemini often brings different searches – and different recommendations.
In your own toolbox
You can copy all queries, websites and brands with one click. If you want to use the sniffer regularly, download it as an LLM skill – a ZIP with a SKILL.md and the resources your own AI assistant needs – or integrate it into your own workflows through the REST API.
Frequently asked questions
Which models can the GEO Sniffer query?
ChatGPT, Claude and Gemini. The same question to different models often brings different searches and different recommendations.
What does the GEO Sniffer cost?
Up to 15 requests a day are free: geotools.hybridai.one.
Why do “Asked” and “Live” differ?
What a model says about its search is not the same as the searches it really sends. For GEO, the real ones count.
Can I use it in my own workflows?
Yes, as an LLM skill for your own AI assistant or through the REST API.
In short: An AI agent that sends emails or changes records on its own needs more than logs: you have to be able to follow every run (turn) and verify that the record of it is complete. Agent observability on the hybridAI platform records every turn with all tool and model calls, secures the events against gaps and tampering with sequence numbers and a hash chain, and raises alerts on anomalies. It is available for hosted HybridClaw agents.
Classic logging
Agent observability
What gets recorded
single events with a timestamp
the whole turn: model, tools, arguments, results, order
Missing data
goes unnoticed
shown as a gap, never reconstructed
Completeness
cannot be verified
sequence number and hash chain, status “broken” with a reason
A chatbot that gives a wrong answer is an annoyance. An agent that sends an email on its own, changes a record or drafts an offer is something else. It acts — and anything that acts has to be able to account for itself.
Most teams putting agents into production discover this at the first incident. The question lands on the table: what exactly happened here? And the honest answer is, surprisingly often, that nobody can say any more.
Why logs are not enough for agents
Classic logging was built for software that executes instructions. One line per event, timestamp, severity level. For a web server that is appropriate.
An agent works differently. Given an input, it decides for itself which tools to use, calls one, judges the result, possibly calls another, and eventually formulates an answer. We call that sequence a turn. What matters is not the individual event but the connection between them: which model was involved, which tools were called with which arguments, what came back — and at what point the decision was made that later turned into the problem.
A chat transcript shows only the surface of that: question in, answer out. Everything in between — which is precisely where the acting happens — stays invisible.
Turn tracing: the sequence as the unit
So we built observability around the turn rather than around the log line. A recorded turn opens up like a protocol: every tool call with its arguments, every result, every model call, in the order it actually happened.
That makes a practical difference. Faced with “why did the agent contact this customer?”, you no longer reconstruct a probable story from scattered log lines. You look at the sequence.
One detail we insisted on: the system does not fill gaps. Where execution history is missing, it says so rather than substituting a plausible-looking reconstruction. Evidence that guesses at the uncertain parts is worse than none, because people believe it.
Completeness you can verify
An agent runs in its own container, often on the customer’s side, and reports its events back to the platform. That raises an uncomfortable question: how do you know nothing was lost on the way — or altered afterwards?
Every event therefore carries a running sequence number and a hash that incorporates its predecessor. On arrival the platform checks both: does the number follow the previous one without a gap, and does the hash match the chain so far? If not, the state is marked broken, with the reason recorded.
This is deliberately stricter than monitoring. Monitoring tells you whether a system is running. A verified event chain tells you whether the record of what it did is complete. For a system that writes into customer data on its own, that is the difference between a log and evidence.
Noticing before someone asks
The same data feeds alerts. Four types cover most of what matters in practice:
Agent offline — the agent has stopped reporting. For a system meant to work in the background, this otherwise surfaces only when a result fails to appear.
Spike in denied actions — the agent repeatedly attempts something it lacks the rights for. Either its task is scoped wrong, or it is heading somewhere nobody intended.
Spike in errors — the failure rate rises against normal operation.
Chain integrity broken — the verification above has failed.
Each alert carries a severity from low to critical and goes to the dashboard, to email, or to a webhook. The point is not the notification itself but that the anomaly is derived from the agent’s behaviour rather than from system load.
See it in the interface
This short video walks through what it looks like in practice — the dashboard, turn search, and the detail view of a single run:
Why this is not a side issue
There is an operational reason and a regulatory one.
Operationally: an agent whose behaviour you cannot follow cannot be improved. You can switch it off or keep trusting it, but you cannot fix the specific point where it went wrong — because you do not know where that point is.
Regulatory: the AI Act requires logging and human oversight for systems at the corresponding risk level. “We have the chat transcript” does not carry that weight. A verified, gap-free record of which tools were called with which data comes considerably closer.
Our position on this is inconvenient but we think it is right: if you deploy agents that act independently, you take responsibility for that action. Observability is the instrument that makes the responsibility exercisable in the first place. Without it, all that is left is trust — and for autonomous systems, trust without verifiability is a poor foundation.
Agent observability is part of the HybridAI platform and available for hosted HybridClaw agents.
Frequently asked questions
What is a turn?
The sequence in which an agent responds to an input: it picks tools, calls them, evaluates the results and formulates an answer. A recorded turn can be expanded like a protocol.
How do I know events are missing or were changed?
Every event carries a running sequence number and a hash that includes its predecessor. If one does not match, the chain is marked “broken” – with the reason.
Which alerts are there?
Agent offline, spike in denied actions, spike in errors and chain integrity broken, each with a severity from low to critical – in the dashboard, by email or to a webhook.
Which agents does this cover?
Agent observability is part of the hybridAI platform and available for hosted HybridClaw agents.
In short: CRM data is incomplete because field sales do not fill in forms after a meeting – not because analysis is missing. The fix is voice input right after the conversation. The hybridAI Sales Companion is an iPhone app with CarPlay that sits as a voice interface on top of your existing CRM – Salesforce, SAP, Dynamics or Zoho. One spoken paragraph becomes a visit report, an offer draft and a deal assessment.
Task
In the CRM today
With the Sales Companion
Visit report
eleven form fields in the evening
dictated in the car park, fields filled automatically
Offer
typing in quantities and prices
spoken → draft, calculated exactly by code, not by the language model
Briefing before the meeting
searching the CRM at your desk
asked on the way: open offers, recent orders, notes, emails
Analysis
waiting for a BI report
in seconds by asking (text-to-SQL)
Data
new system, migration
stays in your existing CRM
Every company running a CRM hears the same complaint from sales: “This system costs me more time than it gives back.” The usual responses are training, mandatory fields, reminder emails, and eventually pressure from management. They rarely work, and the reason is structural.
Why CRM records are incomplete
A field rep has a forty-minute meeting at a customer site. Inside it: a quantity, a reservation about price, the news that the actual decision-maker just changed, a complaint about the last delivery, and a sense of how likely the deal is to close.
How much of that reaches the CRM depends on whether that person — two meetings later, on a train, or at home in the evening — can summon the energy to fill in eleven form fields. Usually one sentence makes it. Sometimes nothing does.
This is not a discipline problem. It is an interface problem. The conversation is spoken language: unstructured, full of context. The CRM expects form fields. Between the two sits a translation job someone has to do by hand — and it falls to the person whose time is the most expensive in the company.
Most of the AI conversation over the past few years has focused on the opposite direction: better analysis of the data that is already there. Dashboards, forecasts, lead scoring. None of that is wrong, but it treats the symptom. If half of what happens in the market never enters the system, even the best model on top of it is analysing the gaps.
Why realtime voice is different from chat
Voice control for business software has existed for twenty years and was almost always disappointing. What changed is not speech recognition — that was already good enough — but what happens between recognition and execution.
A realtime voice model does not listen and transcribe. It understands a spoken paragraph, identifies several distinct facts inside it, routes them to different target structures, and asks when something is missing. So
“Just left Müller GmbH. They want 200 units at the Q3 price, decision by end of month — I’d put it at 70 percent.”
becomes three separate records: a visit report, an offer draft, and a deal assessment. Each lands where it belongs, with the right fields filled in.
Timing is what makes it work. This happens in the car park, two minutes after the meeting, while the details are still sharp — not in the evening, when the memory has worn down and the motivation is gone.
The second difference from chat is that it is hands-free. A rep driving between appointments cannot type, but can talk. And that window — between two meetings — is exactly when people are most willing to deal with something that otherwise gets postponed indefinitely.
Reading and writing belong together
Dictation alone would only be half a solution. It gets interesting when the same voice also answers questions.
“What do I need to know about this customer?” on the drive over — and the answer covers open opportunities, recent orders, notes from the last visit, and the relevant emails from recent weeks. Same data as on the desktop, through an interface that works in a car.
Technically these are two different jobs. Questions about revenue, quantities and pipeline have to become exact queries over structured data — nothing may be estimated, numbers are numbers. Questions about background, conversation history or product detail have to run semantically over unstructured documents: meeting notes, emails, spec sheets, contracts.
We deliberately split this into two paths — text-to-SQL for the figures, semantic search for the story behind them — and let a third model combine and explain the results. The arithmetic is never done by the language model, always by code. An offer whose total an LLM “estimated” would be worthless.
What this looks like in practice: Sales Companion
That is what we built the Sales Companion for: an iPhone app with CarPlay support that sits on the existing CRM as a voice interface.
The Sales Companion voice view: ask a question instead of hunting for a form.
Worth stressing: it is not a new CRM. The data stays in Salesforce, SAP, Dynamics or Zoho, where it already lives. A layer goes on top, reaching those systems through adapters. No migration, no parallel system, no retraining for the teams who keep working at their desks in the tool they know.
Day to day, that means:
Briefing before the meeting — what matters about this customer, summarised on the way there.
Visit report after it — dictated rather than typed, with customer, contacts, topics and next steps.
Offer drafts — spoken quantities and prices become a draft, calculated exactly.
Analysis on demand — “Which customers ordered less this quarter than last?” in seconds, without waiting for a BI report.
One part surprised us with how well it lands: the same voice that knows your customers is also good for practice. A roleplay where the AI plays the counterpart, raises objections and pushes back — followed by a score per skill and concrete advice on what to work on. For new sales hires that beats any handbook.
Privacy is not a footnote
If you send customer conversations through a language model, you need to know where that data goes. Our infrastructure runs on European servers, GDPR and AI Act compliant. Personal data can be masked before a prompt reaches a model, and role-based access controls which teams can reach which data sources at all.
This is not a feature added later. For a system that records visit reports about real people, it is the precondition for being allowed to deploy it.
Frequently asked questions
Is the Sales Companion a new CRM?
No. The data stays in Salesforce, SAP, Dynamics or Zoho; the Sales Companion sits on top through adapters. No migration, no parallel system.
Does the AI calculate offer totals?
No. Numbers come from exact queries on the structured data, and the arithmetic is done by code. An estimated offer would be useless.
Where does the data live?
On European servers, GDPR and AI Act compliant. Personal data can be masked before a prompt reaches a model, and roles control access.
Can I try it?
Yes, on our Agentic CRM page: demo videos and the voice interface on sample sales data, no sign-up required.
See for yourself
On our Agentic CRM page you can watch the demo videos and try the voice interface directly — the public demo runs on sample sales data, no sign-up required.
The more interesting question was never whether the technology works. It is whether your sales team will use it. Our experience so far: when the input takes two minutes in the car park instead of twenty minutes in the evening, they do.
Manage Consent
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behavior or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.