If we contact your company about licensing its data, we're asking you to take an unfamiliar market seriously. We're also asking you to give us your time and, eventually, to trust us with a conversation about records built up over years. That's a lot to ask of somebody who has never heard of Avelence.
Especially now. A business owner may already have emails promising six-figure payments, invoking the names of AI labs and imposing deadlines before anyone has asked what records the business holds. Sometimes the person pitching isn't the buyer. Sometimes they haven't confirmed that the buyer wants this company's data. Sometimes what sounds like a deal is no more than a referral route.
An owner shouldn't have to play detective to work out who's actually on the other end of the conversation. Nor should a serious buyer spend its time working through unqualified introductions made under somebody else's name.
We can't undo that with a claim that we're one of the good ones. We're a newer business too. What we can do is explain exactly where our work begins, where it stops, and what we are prepared to put our name to.
We are an intermediary. We aren't pretending to be the buyer.
Avelence helps companies explore commercial opportunities for the operational data they already hold. We research the buyer market, assess where an archive might fit, help define licensing opportunities, and arrange or support conversations with relevant counterparties. Where an owner wants us involved in negotiations, that role and its authority need to be agreed.
We do not purchase company datasets for our own account. We do not take title to them, collect them into a warehouse, or resell them to AI developers. We are not an AI lab. And we don't run the technical operation that extracts and anonymizes a company's records for delivery.
There are businesses that do those things, including businesses we may work alongside. Some are very good at them. What matters is that a company knows which one it is dealing with.
If we introduce you to a firm that acquires and processes data, we should say that it is the acquiring and processing firm. If we're working under a referral arrangement, that isn't a reason to describe ourselves as its purchasing department. If we are researching potential buyers for an owner, those names are prospects, not confirmed customers for the owner's archive.
We don't need to inflate our position in a transaction to make our work valuable. Finding the right counterparty, understanding what they need, protecting an owner's commercial options and helping structure a deal are substantial jobs on their own.
Knowing a buyer is not the same as having a mandate
This is one of the distinctions that gets lost most often.
A company can publish that it is interested in certain types of enterprise data. Someone can know a person at that company. A referral partner can be authorized to submit qualified businesses. A sourcing team can be asked to find records for a particular task. A buyer can issue a written proposal.
Those are five different situations. They should not be sold to a data owner as though they mean the same thing.
We may identify an attractive buyer route before we have confirmed any interest in a particular archive. When that is the case, we'll say so. If a counterparty has described a specific requirement, we'll distinguish that from a general willingness to consider submissions. If there is a proposal, we'll distinguish its committed terms from amounts contingent on acceptance, delivery, usage or future licensing.
We won't claim that a laboratory has allocated money to buy your company's records simply because we know the market it operates in. We won't borrow a well-known company's name to make an introduction sound more advanced than it is. And we won't put an invented deadline on an opportunity to force a decision.
This may mean some early conversations are less exciting than the pitches circulating online. It also means we can spend those conversations finding out whether there is a real business case instead of defending an exaggerated opening claim.
Your company is not a lead to be passed around
A founder who agrees to talk about their data has not agreed to be circulated through every buyer and intermediary in someone's network.
Before an identifiable opportunity goes out, there should be a reason for approaching that counterparty and permission to share the relevant details. Even an apparently anonymous description can identify a company if it includes a distinctive industry, operating history, product and combination of systems. That deserves the same care as naming the company outright.
We start by understanding who the owner already knows, which conversations may be underway and which parties should be excluded. Then we can work out where a new introduction would be useful. We don't want an owner receiving three approaches to the same buyer under three different referral arrangements. We don't want a buyer seeing the same business presented repeatedly as a new opportunity.
There is a simple commercial reason for this, beyond confidentiality. Buyer relationships are finite. So is a founder's patience. Sending an unqualified proposal to everyone is a good way to use up both.
Our job is to make introductions that stand a reasonable chance of going somewhere, and to be honest when the answer is that we haven't found that fit yet.
We don't need your archive to have the first conversation
The first conversation should be about the business, not an upload link.
We can learn a great deal from the systems a company uses, the years of history it retains, the kinds of work that history records and the authority of the person we're speaking with. Support tickets, engineering issues, CRM activity, project decisions and internal knowledge can all tell different stories. At that stage, we're trying to understand whether any of those stories could be useful to a buyer.
We do not need administrator credentials, a full Slack export, customer conversations or the entire source-code repository to answer that initial question.
If a specific opportunity warrants deeper diligence, the next stage needs to be agreed with the owner. Which records? Who may inspect them? What can safely be disclosed? What has to be excluded? Who is responsible for preparing a sample? Where will it be processed and under whose controls?
A confidentiality agreement can help protect a discussion, but it doesn't answer all of those questions. An NDA does not turn a customer's confidential information into the company's own intellectual property. It doesn't give an employee authority to sign for a business. And it doesn't turn permission to inspect a sample into permission to train a model on the entire archive.
We won't ask an owner to confuse those steps. We also won't pretend that Avelence is providing technical security, data removal or anonymization services that another party would actually have to perform.
Possessing the records doesn't mean you can license all of them
This matters particularly for the kind of data we work with, because useful operational history tends to contain more than one party's information.
Take an old support ticket. The investigation and resolution may be valuable examples of how a company solves problems. The ticket may also contain a customer's confidential configuration, someone's personal details, a vendor's documentation or an internal security discussion. The fact that the ticket sits in the company's helpdesk doesn't settle which parts it is entitled to license for AI development.
The same applies to code repositories, sales histories, employment records, contracts and internal communications. A business may have created part of the material, received part of it under a restricted agreement and retained another part only for a particular purpose.
We want to surface those distinctions before anyone promises a dataset. Sometimes that means narrowing a proposed scope. Sometimes it means excluding an entire source system. Sometimes it means the owner and its advisers need to establish rights before a buyer can proceed.
Removing names is not a substitute for that work. De-identification can reduce exposure, but linked events, unusual facts and retained context may still reveal people or confidential business information. The answer depends on the records and intended use, not the name of the tool used to process them.
If the rights and handling cannot be established on acceptable terms, we would rather see the opportunity stop than ask an owner to make promises it cannot safely keep.
A large archive is not automatically a valuable dataset
There are good reasons a buyer may want historical enterprise records. Models intended to do real work need to understand the work itself: the problem that arrived, the information available, the steps taken, the decisions made and what happened as a result.
But volume on its own tells us surprisingly little.
Consider a software company with years of support tickets, engineering issues and code reviews. One buyer may be interested in cases where a support problem can be traced to a technical investigation, a change in code and a verified resolution. Another may be looking for repeated operational workflows and updates over time. The same company could be relevant to both, but not with the same sample, evaluation criteria or licensing terms.
A folder full of disconnected tickets might be much less useful for the first task than a smaller set of complete incident histories. A technically rich archive might be unsuitable if the needed permissions are missing. A valuable source may also require significant preparation before a buyer can assess it.
That's why our work starts with the buyer's use case and what the records can actually demonstrate. We aren't selling file counts, terabytes or years in business as proof of value. We are trying to establish whether there is a specific fit worth developing.
For buyers, the standard works in reverse. Tell us the task, the required record structure, the exclusions and how you intend to evaluate a source. We can research and qualify possible companies against that brief. We shouldn't send an unverified owner with a compelling story and ask your team to discover the missing requirements for us.
We won't give you a price and pretend it's an offer
We have a valuation calculator on our homepage, and we intend to keep it.
It serves a useful purpose. Many business owners have never considered that their operating history might have commercial value beyond their own company. A numerical range gives them a reason to look more closely and ask the next question.
The range brings together selected public pricing models from established data-buying and licensing businesses. It is a starting estimate, not a commitment from any of those businesses. No one has evaluated your rights, the completeness of your records or the demand for your particular archive simply because you've entered your company's size and history.
We also won't use the largest number someone has ever advertised as if it were a normal outcome. A payment headline can hide quite different arrangements: money paid up front, payment after acceptance, revenue contingent on future licensing, obligations to deliver updates or rights granted exclusively for years.
Those differences can change the value of a deal substantially.
We want owners to compare what they would actually receive and what they would give up. If a buyer offers more money in exchange for exclusive use, the comparison isn't finished until we've considered what that exclusivity prevents. If a recurring arrangement has a higher theoretical ceiling but no guaranteed minimum, it shouldn't be described as equivalent to a committed payment.
The calculator opens the conversation. The buyer research, commercial structure and negotiations are where a real valuation begins to take shape.
We don't confuse being in the middle with owning the transaction
There is nothing inherently wrong with being paid to source a company, make an introduction or advise on a licensing negotiation. Those functions exist in plenty of mature industries. The fee isn't the problem. Hiding what the fee is for, or implying authority that doesn't exist, is the problem.
Our involvement needs clear boundaries. Who engaged us? Whose interests are we representing in this stage? Are we providing research, arranging an introduction, supporting negotiation or leading it under an agreed mandate? Who would pay us, under what terms, and is there a referral relationship the parties ought to understand?
An introduction does not automatically give an intermediary control over every future discussion between two companies. Being paid by a sourcing partner does not make us an employee of a buyer. Advising an owner does not give us permission to sign its licensing agreement.
Some owners will want us to handle much of the commercial process. Others may want a buyer shortlist, an introduction or a second view on an offer before taking discussions forward themselves. Either can be a sensible engagement. We should agree on the work and economics rather than make the owner discover them midway through a negotiation.
We also have no reason to turn a one-off introduction fee into a claim about annual recurring revenue, or to present the total value of a licensing contract as if it were Avelence's own earnings. Those are different amounts, and anyone doing this work should know the difference.
The buyer deserves the same straight answers
A legitimate buyer has work to do beyond finding somebody who says they own a lot of data.
The buying team needs to know what task a dataset might support, where the records came from, whether the seller can authorize the proposed use, what the missing or unusable portions look like, and who will be responsible for preparing and delivering an acceptable contribution. It also needs to understand whether the source is being represented by several intermediaries and whether prior commitments limit what can be licensed.
We don't gain anything by hiding those questions until after an introduction. An evaluation that fails because the owner was never authorized, the records don't connect or the same corpus was promised exclusively elsewhere costs the buyer time and costs the owner goodwill.
We may not have every answer at first contact. That's normal. The important thing is to distinguish a verified fact from a question still being investigated. We would rather say the owner has eight years of records, but the linked outcomes need to be checked than call the dataset model-ready because the volume sounds impressive.
We don't claim to have a prepared inventory of data sitting behind our website. Our buyer-side service is to understand the requirement, identify plausible owners, qualify the opportunity and coordinate the next conversation under the right authority.
Sometimes the right answer is to leave the data where it is
We are building a business around the fact that enterprise records can have value to AI developers. That doesn't mean every archive should be licensed.
A company may have useful records it cannot lawfully or contractually share. The buyer's use may be too broad. The preparation cost may erase the upside. An exclusive agreement may close off better future opportunities. A founder may reasonably decide that helping train a system on a core internal process works against the company's interests.
Those are commercial decisions, not objections to be talked around until someone signs.
We can help an owner narrow a proposed license, consider different buyers or decide to revisit the market later. We can also tell a buyer that a source doesn't meet the requirement. A relationship that ends with a clear no is still preferable to one built on promises that unravel after weeks of diligence.
This is especially important now. If an owner has already sat through several exaggerated pitches, the next conversation is starting with a trust deficit. We won't repair that by promising to be different. We can only repair it by being specific, respecting the answer and doing what we said we'd do.
What you should expect from us
Avelence is a commercial partner in enterprise AI data sourcing and licensing. We help owners understand which parts of their operating history might matter to buyers, how to approach the market, and how to compare the terms available. We help buyers find and qualify companies that may hold data suitable for a defined requirement. We work with specialist intermediaries and technical providers where their services are needed and the roles are understood.
We start with a high-level description, not a demand for raw files. We make the difference between possible demand and confirmed interest clear. We establish permissions for introductions and deeper disclosure. We are explicit about the type of engagement and the authority it gives us. We assess the commercial substance of an offer instead of stopping at its headline number.
Some of those things sound basic. They are basic. That is precisely why we think they should be said plainly.
We are a young company. We aren't claiming a history of transactions or laboratory mandates that we cannot substantiate. We expect to be judged on the quality of our work, the relationships we build and how accurately we represent each opportunity, including the ones that go nowhere.
If you are going to trust us with a business relationship, these are fair things to ask us to prove.
Questions worth asking us
Are you a data buyer or an AI data broker?
We are not a purchaser of datasets. Avelence is a sourcing and commercial advisory intermediary. We help owners identify potential buyers and structure licensing opportunities, and we help buyers find and qualify relevant companies. We do not buy, warehouse or resell the underlying records. If someone uses the word broker to describe an intermediary, the more important question is what the intermediary is actually authorized to do.
Are you working directly for an AI lab?
We will describe the actual counterparty and our role in a specific opportunity. Knowing people in the market, researching a buyer or participating in a referral programme does not amount to a purchasing mandate. We will not claim to represent a laboratory unless we have that authority. Some buyer identities or requirements may be confidential at an early stage; we still have a responsibility not to misrepresent what has been confirmed.
Can you guarantee a payment for our company data?
No. Avelence does not issue purchase offers for datasets. Our homepage calculator gives an indicative range built from selected public pricing models, not a guaranteed price. Actual commercial terms depend on the buyer, the records, their permitted use, preparation and acceptance requirements, and negotiation. We distinguish an estimate, an expression of interest, a written offer and money received.
Will you ask us to share our raw company records?
Not to begin. A useful first conversation usually needs only a nonconfidential picture of the business, its systems, the history it retains and the person authorized to discuss it. If a buyer later requests a sample or a delivery, the owner and the relevant technical and legal parties need to agree the scope and handling before anything moves.
Who gets paid if Avelence introduces us to a buyer?
That depends on the specific advisory, sourcing or referral arrangement. We should agree on our role, who pays us, what earns a fee and any material conflict before the parties rely on that involvement. A buyer's total contract value is not automatically Avelence's revenue, and an introductory conversation doesn't, by itself, entitle us to a fee on every future deal.
Can we approach a buyer ourselves or decide not to proceed?
Yes. Owners can choose to contact buyers directly, subject to any agreements they have knowingly made, or engage us for a defined part of the commercial work. Exploring the market does not oblige a company to license its records. If the rights, terms or strategic implications don't make sense, walking away may be the right decision.
We're not asking you to believe a story about the AI data market. We're asking you to judge the work in front of you.
If we recommend a buyer, you should be able to ask why. If we suggest a licensing structure, you should be able to understand what you gain and what you surrender. And if there is no credible deal to pursue, you should hear that before you share an archive or spend weeks on calls.
That's the standard we're putting on the record. Hold us to it.
Founder & CEO, Avelence