Tasken AI

BACKGROUND

Please describe your company and position.

I am the Head of Outreach of Marketing Signals

Describe what your company does in a single sentence.

Marketing Signals is a digital marketing agency that delivers SEO, AI search visibility, PPC, Digital PR and link building

OPPORTUNITY / CHALLENGE

What specific goals or objectives did you hire Tasken AI to accomplish?

• Replace manual site auditing with AI-driven domain qualification at scale.
• Automate outreach, from contact finding to campaigns and reply handling.
• Process large lists quickly at a lower cost than our in-house team.

SOLUTION

How did you find Tasken AI?

Referral

Why did you select Tasken AI over others?

Referred to me

How many teammates from Tasken AI were assigned to this project?

2-5 Employees

Describe the scope of work in detail. Please include a summary of key deliverables.

Tasken AI built us a custom automation system to take over most of our link-building workflow, which our back-office team had been doing by hand. They worked from December 2025 to March 2026 on the initial build and stayed on afterwards via a support package to iterate on the system. The main deliverables were:

Domain qualification (Website Audit): We upload lists of thousands of domains. The system filters them by TLD, Trust Flow and organic traffic, then uses AI to check each site's language, niche, activity and quality, and scores and classifies it. Tasken rebuilt the backend so that 1,000 domains take 20–30 minutes instead of around 9 hours. They also tested AI models to bring the cost per 1,000 domains down to a workable level.

Contact discovery and outreach: The system scrapes and verifies publisher emails, with a fallback provider for sites where no email is found. It connects to our Instantly account, so we can create campaigns, write personalised opening lines and edit sequences without leaving the dashboard.

Reply handling: Publisher replies come into an inbox inside the tool. They're sorted by type (interested, price quote, no), and a draft response is written for each one.

Link insertion: The system scans a publisher's existing articles, scores each one on relevance, traffic and how easily a link could fit, then suggests anchors and drafts the outreach email.

Platform upgrades: Tasken moved the original Softr/n8n build onto a faster, custom-coded version 2. It added AI search across the publisher database, link pricing pulled from marketplaces, client relevance scoring, and prospecting via Google and marketplaces.

Handover and support: Tasken set up a Claude project with the system's documentation so we can refine prompts ourselves. They gave ongoing maintenance and prompt tuning, and are moving the system onto our own infrastructure (Trigger.dev, Supabase, GitHub).

RESULTS & FEEDBACK

What were the measurable outcomes from the project that demonstrate progress or success?

Speed: The time to audit 1,000 domains fell from about 9 hours to 20–30 minutes. Per domain, that's 1–3 seconds instead of 1–2 minutes.

Cost: Early builds cost about $40 per 1,000 domains in AI usage (over $100 before Tasken optimised it). After model testing and prompt work, our own OpenRouter spend came to about $11.50 per 1,000 sites.

Scale: The system now holds more than 52,000 audited domains and returns filtered results in under a second. It handles batch uploads in the thousands, which our back-office team used to process by hand.

Output quality: In one 2,000-domain test run, the system qualified 703 unique domains. Our outreach team reviewed several test batches and found the results looked good.

Describe their project management. Did they deliver items on time? How did they respond to your needs?

Tasken were responsive throughout the project. We worked with them in a shared Slack channel, they usually replied the same day, and they posted regular progress updates. Whenever we raised a problem, such as stuck batches, AI costs or slow processing, they looked into it quickly. Often they came back with a fix or a new approach within a day or two. Their cost work is a good example. They tested several AI models and brought our cost per 1,000 domains down to a level that made the tool viable for us.

Overall they were proactive and flexible with us. They honoured the pricing we originally agreed, offered to move the system onto our servers at no extra cost, and flagged problems like low API credits before they became blockers.

What was your primary form of communication with Tasken AI?

• Virtual Meeting
• Email or Messaging App

What did you find most impressive or unique about this company?

The most impressive thing was how hard they kept working on the product after the initial build. They didn't hand over a tool and move on. They kept improving it: they cut processing time from hours to minutes, tested AI models until running costs worked for us, and added features we hadn't asked for, such as reply handling, link pricing and AI search across our publisher database. 

They also set us up to manage the system ourselves. They built a Claude project with the full technical documentation, so our team can refine the prompts without depending on them. They're now moving everything onto our own infrastructure.

Are there any areas for improvement or something Tasken AI could have done differently?

Explaining running costs upfront. We didn't fully understand the ongoing AI and API costs until testing had started. In the early builds, those costs made the tool more expensive than doing the work manually. Tasken solved this in the end, but knowing earlier would have helped us plan.

Communicating changes. Some updates were made without being agreed with us first. When they moved to version 2, for example, we lost access to the old system and our historic data for a while without warning. Agreeing changes in advance and announcing migrations clearly would have made things smoother.

However, in both instances, Tasken moved very quickly to make amends.