How Tendem works: from goal to finished result

A place to send the work that still needs finishing

A draft that takes AI a minute to produce can still leave you with an afternoon’s work. The sources need checking, company details are missing, or the contacts don’t quite match the customers you had in mind. Before long, you’re spending the time you thought you’d saved.

For founders and lean teams, that work often lands with the same person. Market research, lead-list building, and outreach writing need different skills, but hiring someone for each job would become a project of its own.

Tendem is an on-demand expert bench that returns finished work. Bring a goal, a brief, a file, or a draft you’ve already started. Tendem agrees on what you’ll receive and the price, runs the work through the right mix of AI, vetted experts, and quality checks, then brings the result back for review.

You set the outcome. Tendem manages the steps needed to get there.

Who’s finishing the job?

Say you need a list of potential customers for a new campaign. You know the kinds of companies you want to reach, but you have never built a lead list properly before.

Doing it yourself means deciding how to find companies, identify the right people, fill in missing details, and verify the data. Buying a list means choosing a provider whose quality you may not know how to evaluate. An AI tool can produce a spreadsheet quickly, but missing fields and uncertain matches can leave you checking it row by row.

The work includes judgment as well as finding information. Does this company really fit? Is the person still in the relevant role? Was the email checked, or simply found somewhere online?

With Tendem, you can hand over the complete task: define the target audience, agree on the fields and checks, and receive a structured list evaluated against that brief. Tendem manages the research, enrichment, and applicable verification steps.

A freelancer or agency can also help with this work. Tendem gives you a way to commission a defined result without sourcing the individual contributors or coordinating their work yourself.

What can you hand off?

Lead lists are one example. Market research is another: a plausible answer still needs sources, clear boundaries, and a method that fits the question. These are some of the jobs you can scope with Tendem:

For example, a founder exploring a new market could start with a research question, then commission a list of relevant companies, and later ask for outreach copy. Each can be submitted as a separate task. If several are needed together, their deliverables and relationship should be included in the scope.

Research a market when the answers don’t agree

Suppose you want to understand the US market for customer-onboarding software serving B2B software companies with 100 to 1,000 employees. Several AI estimates might look plausible while counting different products, buyers, or spending assumptions.

A useful brief would define the decision the research supports, include dedicated onboarding platforms, and exclude general project-management tools and product-adoption widgets. You could ask for a market range, a bottom-up calculation, dated sources, assumptions, and evidence that challenges the conclusion.

The result should make the reasoning inspectable. You can see what was counted, how the estimate was built, and where it remains uncertain. The task is scoped around an answer you can assess and use in a strategy discussion.

Write outreach once you know whom to contact

A qualified list still leaves the question of what to say. A polished message can lead with the wrong buyer problem, make an unsupported promise, or repeat the same point across every follow-up.

For an outreach task, bring the offer, customer profile, real customer evidence, and voice guidance. Specify which claims are permitted, what language to avoid, and which channels need copy.

The agreed package could include email and LinkedIn sequences, distinct messaging angles, and guidance on what to personalize. With B2B copywriting expertise included in the task, the work can address the opening argument, supporting proof, and purpose of each follow-up as well as the wording. Checks cover those requirements before delivery. You decide whom to contact and how to run the campaign.

What happens after you send a task?

1. Describe the result you need

Start in the Tendem web platform with a plain-language request, or choose a task from the template library.

A request might look like this:

Build a list of 100 potential customers matching the attached ideal customer profile. Include company, contact name, role, profile link, work email, and verification status. Exclude existing customers and anyone on our suppression list. Return a CSV using the attached column names.

This is an illustrative brief. The target quantity, required fields, and treatment of unavailable data are agreed during scoping.

You don’t need to design a research and verification workflow. Give Tendem the goal, the types of customers you want to reach, and the material it can use. If there is a deadline or budget to work within, include that too.

Templates provide a more structured starting point for familiar jobs. They set out the information needed, the expected steps and checks, and the finished deliverable. You add the specifics of your task. If nothing fits, describe the work yourself.

2. Fill in the details

Tendem asks questions to turn the request into a clear scope. For the lead list, that might mean confirming geography, company size, target roles and seniority, and what makes a company a good fit. A few example companies can help clarify the audience.

The scope should also establish which fields are essential, which verification statuses are acceptable, and what happens if fewer than the target number meet the criteria. Filling the spreadsheet with weak matches would not meet the same requirement.

The scope records the objective, required deliverables, format, and relevant boundaries. It can also specify the sources, assumptions, exclusions, and evidence the work needs.

This is the point to make preferences concrete. Share example customers, identify companies that must be excluded, and provide the column structure needed for your CRM. Those details give both the people doing the work and the checks a useful standard to follow.

3. Review the scope, price, and estimated turnaround

Before work starts, review what will come back, what it costs, the estimated completion time, and the expected expertise and checks.

Tell Tendem which constraint matters most, whether that is budget, timing, or particular expertise. Tendem works out how the task can be carried out within the agreed scope. You don’t need to assign individual steps to AI or people.

Human involvement depends on the execution option you approve. Some tasks run on AI with automated checks. Other options include an expert to carry out work, guide it, or review it. A human does not work on every task.

Every task has a visible price, including when a plan covers it. Pay-as-you-go is available, and subscriptions provide another way to pay over time. A plan determines payment and coverage; the task’s agreed execution option determines how the work is done. Current plans and terms are on the pricing page.

The estimated turnaround applies to the particular job. Checking an existing contact list and researching a new market before building a list have different requirements. Check the estimate against your deadline before confirming the work.

4. Tendem carries out the work

Tendem divides the task into steps and assigns each one according to what it needs. AI and experts can contribute at different points, with some steps running together and others following in sequence.

For the lead list, a specialist might help interpret the targeting criteria before research starts. Company research, contact identification, enrichment, and verification can then be handled as the task requires. Expert involvement can mean doing the work, guiding it, or checking it, depending on the approved task.

Tendem manages those assignments and handoffs. You don’t need to post a job, compare profiles, or coordinate the contributors.

5. The result is checked and assembled

The approved brief determines what the result is checked against. Depending on the task and execution option, checks can cover factual accuracy, sources, completeness, technical correctness, brand voice, design, and usability.

For the lead list, that could mean checking company fit, required contact fields, duplicates, exclusions, and recorded verification status. Finding an email address and verifying it are different steps. A verification status documents a check; it does not guarantee that a future email will be delivered.

For market research, the checks could examine whether sources support the conclusions, the market boundaries are clear, and assumptions or conflicting evidence are documented.

The work comes back as one completed task or project outcome. That may include several files, such as the lead-list CSV and a record of verification or exclusions, if those were part of the scope.

6. Approve the result or request changes

When the work arrives, choose Approve to close the task or Request changes to explain what needs adjusting. The decision covers the whole task, including any files delivered together.

You can review the result before accepting it. If something needs changing, there is a defined process for getting it addressed.

Who checks the work?

The person or AI that produces a piece of work does not approve its own contribution. Checking goes to a different executor. An automated checker starts with fresh context rather than relying on the producer’s reasoning.

The checks also need something concrete to work with. A factual claim may need a supporting source. A calculation may need a documented method and assumptions. Code may need tests. If a check finds a problem, it should identify what failed against the brief.

The exact review depends on the work and the execution option approved. A sourced market report needs different scrutiny from a logo or a software fix. Not every task receives the same checks or human-review layer.

Tendem takes responsibility for the execution, checks, and deliverables promised in the scope. You decide how to use the result, whether that means choosing a market to pursue or deciding whom to contact and how to run the outreach.

The experts behind Tendem

Tendem draws on more than 10,000 vetted experts across 20+ domains, built on Toloka’s long-running expert network and operating experience. Matching considers the skills the task needs and the expert’s performance history.

Ready-made templates draw on task types Tendem already understands, with an established workflow, relevant expertise, and a defined quality bar. That gives familiar jobs a more specific starting point than a blank request.

Tendem’s customers include Microsoft, JetBrains, Poolside, and Hugging Face. For an individual task, the practical reassurance starts before execution: an agreed deliverable, a visible price, defined checks, and a way to request changes after delivery.

What if the result needs changing?

Choose Request changes and describe the issue in plain words. Tendem asks clarifying questions when needed and compares the feedback with the scope you approved.

Corrections and small adjustments to the same deliverable can be handled as a free redo. For example, you might point out contacts outside the agreed company-size range or missing verification statuses that the brief required. Asking for another list in a new geography introduces additional work that needs its own scope and price.

You don’t need to classify the feedback yourself. One message can include an error, an adjustment, and an additional request. Tendem separates those into the appropriate next steps.

For a redo, Tendem shows the updated scope. Review it and select Start. The work continues within the same task, without another offer-selection process. Each completed redo produces a new version, and earlier versions remain available.

The policy provides three free change attempts per task, shared between redos and small adjustments. Remaining attempts are shown in the product and stated in the task terms. A new task does not use an attempt; cancelling a redo after it starts does.

If those attempts run out, the next step depends on what remains unresolved. Expanded requirements may need a new paid task. If Tendem has failed to deliver the agreed scope, support reviews the case and may issue a credit refund. That decision is made by support and is not automatic.

Using Tendem inside your AI tools

You can also call Tendem from a compatible AI tool or agent workflow through MCP, the connection that lets the tool pass work to Tendem and receive the result back.

For example, an agent building a company list might reach records it cannot verify. It can send that work to Tendem with the relevant context. Tendem scopes the task and presents the price before execution. After approval and completion, the result returns to the original workflow as markdown and files, so the agent can continue.

Compatible clients include Claude, ChatGPT, Cursor, and Copilot, using a custom MCP connection and a Tendem account. Agent pipelines can connect to the remote MCP endpoint with an account API key.

This route is useful for verification, specialist assessments, missing data, and reviews of work an agent has started. See the Tendem MCP guide for connection instructions and supported setups.

Bring one task to Tendem

A useful first task is a piece of work with a clear destination: a sourced market assessment for a planning meeting, a competitor comparison, or a checked lead list for your CRM.

Bring the material you have and describe what needs to come back. Tendem helps establish the scope, shows the price and estimated turnaround, and manages the work through to a result you can review.

Start a task or choose a template.

The most popular tasks on Tendem

1. Research a market before committing to it

Starting point: A founder has a market question and several plausible AI answers, but no clear basis for choosing between them.

Example brief: “Estimate the US market for customer-onboarding software serving B2B software companies with 100 to 1,000 employees. Focus on dedicated onboarding platforms and exclude general project-management tools and product-adoption widgets. Return a defensible range for a strategy discussion, with a bottom-up calculation, dated sources, explicit assumptions, competitor boundaries, and counter-evidence.”

What comes back: A sourced analysis with market boundaries, methodology, assumptions, limitations, and relevant counter-evidence. Checks focus on whether the evidence supports the conclusions. The result supports a strategy discussion; it does not guarantee the business decision.

2. Build a lead list that meets the actual brief

Starting point: A lean sales team has an ICP and a partial spreadsheet with missing fields and inconsistent records.

Example brief: “Build a list of 50 UK B2B software companies that meet the attached criteria. Include company size, website, the relevant marketing contact, source links, and verification status. Exclude current customers and the attached suppression list. Return a CSV using these column names.”

What comes back: A structured list checked against the targeting criteria, with required fields, documented verification status, and exclusions applied. The agreed format prepares it for CRM import without implying a native CRM integration. Email verification does not guarantee deliverability.

3. Write outreach grounded in the offer

Starting point: A founder has a product and a target audience, but the existing messages sound generic or make unsupported promises.

Example brief: “Write a three-email sequence and a separate LinkedIn sequence for this ICP. Use the attached offer and real customer evidence. Lead with the buyer’s problem, use only approved claims, and give each follow-up a new reason to respond. Include distinct angles and guidance on what to personalize. Avoid fake compliments and manufactured familiarity.”

What comes back: Distinct sequences with supported claims, relevant opening angles, and follow-ups that add a reason to respond. Review covers the audience, offer, tone, claims, and required components. The team gets copy to use in its sending tools, without a promise of replies or meetings.

4. Get a logo that works in its intended formats

Starting point: A small business has a name and visual references but needs a designer to turn the idea into usable assets.

Example brief: “Create a logo for this company and audience. It needs to work on our website, a dark background, and as a favicon. Include the source files and one-color variants. Use these references for direction.”

What comes back: The agreed logo files and variants, with designer judgment and checks against the named uses. The scope makes clear which assets are included, so a logo request does not silently become a complete brand identity project.

5. Clean and verify a dataset

Starting point: A team has collected records, but duplicates, missing values, and uncertain sources make the dataset difficult to use.

Example brief: “Clean this supplier dataset. Apply the attached schema, remove duplicates, verify the required fields against the agreed sources, and record verification status. Exclude records that fail the requirements and document the exclusions.”

What comes back: A structured dataset with the agreed checks documented and failed records handled transparently. The team can see what was checked instead of assuming that a populated row is reliable.

6. Fix and test a defined software issue

Starting point: A nontechnical founder or small product team has a reproducible bug and needs technical help to resolve it.

Example brief: “Fix this form-validation bug using the supplied code and reproduction steps. Cover the listed edge cases, add appropriate tests, and return the code changes with a short explanation of the fix and test results.”

What comes back: The scoped code changes and testing evidence, with technical review as agreed. Access requirements, environment, delivery method, and any deployment work should be established before execution.

7. Add expert review to an agent workflow

Starting point: An automated research workflow has produced a draft but cannot confidently judge its technical claims.

Example brief: “Review these claims against the supplied sources. Flag unsupported conclusions, explain the issues, and return corrected statements with references in the specified structure.”

What comes back: An assessment and supporting evidence returned through MCP as markdown and files. The agent can use the result in its next step, with the human expertise and checks defined in the task scope.