Let Herald agents handle your experts' non-billable work.
Herald agents handle the recurring work your experts shouldn't be spending time on.
Heralds is the platform for building and running AI agents for expert work.
Working now
Check sources
- Prepare
- Check sources
- Assess relevance
- Prepare update
- Review if needed
- 15:00:24Check queuedHeralds saved this run and will begin with your confirmed settings.
- 15:00:41Monitoring settings preparedThe source check is starting with the settings saved for this run.
- 15:01:09Checking sourcesLooking for material added or changed since the previous run.
You can leave the page. Every step and result is saved.
§ The drag
The work that never makes it onto an invoice.
In expert businesses, the same manual work repeats every week. It is necessary, it is skilled, and almost none of it is billable.
- Checking sources for anything new
- Reading updates nobody has time to read
- Summarising what changed, and for whom
- Preparing recurring client reports
- Qualifying and researching leads
- Monitoring competitors, regulation and news
- Drafting internal updates
§ Why the AI experiments stall
- 01
No time
Building the workflow competes with the billable work you already have.
- 02
No know-how
Most experts are not prompt engineers, and should not have to be.
- 03
No trust
A confident answer with no sources is worse than no answer at all.
- 04
Nothing repeatable
A good chat session is not a process. Next week it starts from zero.
- 05
No business context
Generic output, because the tool knows nothing about your firm or clients.
§ Foundation
Heralds learns your business first.
Give Heralds your domain. It researches your public footprint and turns it into an editable company profile you approve.
- 01Heralds reads your public pages and proposes facts: what you do, your products and services, your ICP, your languages, your public team.
- 02Every field carries a confidence level and a source link, and facts appear live as they are verified.
- 03You confirm or uncheck each one. Nothing counts as company context until you review it.
- 04That confirmed context prefills every agent you create, which is why the results are about your business and not a generic industry.
Heralds is working
Checking which public pages describe the business
Research steps
- ✓Plan the research
- ✓Read the company website
- ✓Understand the business
- 04Match your public role
- 05Prepare your profile
- What the company doesB2B software for logistics teamsHigh confidence · source linked
- Products and servicesRoute planning · Fleet analyticsHigh confidence · source linked
- Ideal customer profileMid-size carriers in Northern EuropeMedium confidence · source linked
- Your roleHead of GrowthNeeds your review · source linked
Nothing becomes a confirmed company fact until you review it.
§ How agents work
Every agent works the same way.
Scan, analyze, report. Manual or scheduled, and never a black box. Learn it once and every new agent is familiar.
- 01
Scan
Check the sources that matter for anything new since the last run.
- 02
Analyze
Assess every new item against your confirmed business context.
- 03
Report
Show only what matters, with the reasoning and the source links.
Working now
Check sources
- Prepare
- Check sources
- Assess relevance
- Prepare update
- Review if needed
- 15:00:24Check queuedHeralds saved this run and will begin with your confirmed settings.
- 15:00:41Monitoring settings preparedThe source check is starting with the settings saved for this run.
- 15:01:09Checking sourcesLooking for material added or changed since the previous run.
You can leave the page. Every step and result is saved.
While an agent runs you see the stage it is on and a timestamped log of what it did. Leave the page whenever you like, because every step and result is saved.
We keep up with AI, so you don't have to. No lock-in to a single model.
Heralds keeps your agents on current AI models and tools, so your team never has to track the AI market.
§ Agents
Choose an outcome, not a toolchain.
Each agent knows what to watch, what to filter for, and what to produce. If none of them fit, describe your own in plain language.
Sales Lead Agent
Live- Watches
- Company news, funding, hiring and other public buying signals.
- Filters for
- Fit with your confirmed customer profile, and whether the timing is now.
- Produces
- A short, scored lead brief with the signal, the person to talk to and an opening angle.
Removes the manual research pass before every outreach list.
Legal Agent
Pilot- Watches
- Legislation, regulators and official legal sources you select.
- Filters for
- Only changes that touch your practice areas or your clients.
- Produces
- A client-advisory style summary of what changed and who it affects.
Replaces the weekly read-everything routine with a reviewed shortlist.
Reputation Agent
Preview- Watches
- New public mentions, reviews, competitor and market moves.
- Filters for
- Sentiment, reach and whether a response is actually warranted.
- Produces
- A short list of items that matter, each with the source and a suggested next step.
Stops the daily manual scan of search results and review sites.
Custom Agent Builder
Guided- Watches
- Whatever recurring job you describe in plain language.
- Filters for
- The rules you confirm during a short guided setup.
- Produces
- A working agent you review before it runs, manually or on a schedule.
Turns a workflow you repeat every week into something you approve, not perform.
§ Try it
Pick an agent. Watch it work.
Switch between the example agents and see the run and the result change. Every example below is illustrative, but the shape is exactly what you get.
Working now
Prepare
- Prepare
- Check sources
- Assess relevance
- Prepare update
- Review if needed
- 09:00:04Run queuedStarting with your confirmed customer profile and source list.
You can leave the page. Every step and result is saved.
The result appears here when the run finishes, with the evidence and the source behind it.
§ The output
The same report, in any channel.
Every agent reports in the same six parts, in the channel you choose. Nothing reaches you without evidence behind it. If a finding cannot be backed with a source and a reason it matters to you, it is not shown at all.
The evidence
The source material the agent checked, with a link to the original record.
Why now
What changed and why it triggered this alert today.
Relevance
How it matches the context and rules you confirmed.
What it means
The implication for your business and why it matters.
Suggested next step
A concrete action the agent recommends based on the evidence.
Source
Heralds checked the sources you connected.
The same structure is used for a legal update, a sales signal, a reputation alert, or any custom agent you build.
§ Proof
Built by a team that has shipped AI agents before.
Before Heralds, we built and exited Cuckoo, a B2B SaaS company run with AI agents across support, sales, product, QA, security, reporting and internal operations.
We know what it takes to get an agent past the demo stage, and how much of that work should never have been necessary. Heralds is the product we wish we had.
What we learned
- An agent is only as good as the business context behind it.
- People adopt what they can watch and correct.
- Evidence beats confidence, every time.
§ Start with one workflow
Get your first agent running.
Pick one recurring job that never makes it onto an invoice. Tell us where to reach you and we'll set that agent up with your business context.