Every Application Read in Full. Every Score Backed by Evidence.

Q: Does IdeaScore AI decide who gets in?

No. People make every decision. Its score sits beside the panel’s, never blended in, and a helper never changes a score or decides an outcome. Applicants are told in plain words when IdeaScore AI helped.

Q: Which tier should we use?

Swift sorts, Pro reviews, Deep decides. Swift gives a fast first read of a large intake, Pro is the everyday full review with evidence and confidence, and Deep researches every claim and reviews every score a second time for shortlists and final rounds.

Q: How are the tier figures worked out?

Pro 4’s figures are measured against expert reviewers’ scores. Swift 4 and Deep 4.5 figures are estimates, compared with expert reviewers’ scores, and are marked estimated wherever they appear.

Q: What does Deep Research cost?

Deep Research also searches the web as well as your programme, and lists the pages it used. On a chat or Ask question it adds 8 credits on Pro and Deep, and you see the cost before you send. A Pro evaluation is 5 credits, or 15 with Deep Research; a Deep evaluation is 30 credits with the research included. Swift does not offer it.

Q: Are the helpers charged?

No. Helpers are free with fair use. Credits pay for evaluations, chat, Ask, cohort questions and Deep Research, and every helper says in one line what it reads.

Q: What is sent to IdeaScore AI?

Only what a run needs from the application and your programme. Names and contact details are swapped for labels first, even inside answers and attachments, and applications are never used to train AI models.

Q: Can evaluators see IdeaScore AI’s scores?

Not unless you choose. Its notes are hidden from evaluators by default; your organisation can show them after an evaluator submits their own score, or always. Programmes can only be stricter, and blind rounds always hide them.

Q: Can we switch it off?

Yes. An owner can switch IdeaScore AI off for the whole organisation or for a single programme, so a sensitive call can run with people only.

IdeaScore AI reads every application against your rubric, with the evidence beside every score, so your panel spends its time on judgement.IdeaScore AI reads every application in full, scores it against your rubric with the evidence beside every score, and answers anything about your programme, so your panel spends its time on judgement, not first reads.

Book a demo Compare Swift, Pro and Deep

Halcyon Grid #0142

IdeaScore AI · Pro 4

44 / 50

Innovation and noveltyHigh confidence9 / 10
“Our controller balances storage at the feeder head, below 1 MW, where no product serves today.”
Quote checked · Two comparable ventures found; neither targets sub-1 MW feeders.
Customer desirabilityMedium confidence8 / 10
“A signed pilot with one rural co-operative; a second utility is in discussion.”
Quote checked · A second signed pilot would move this up a band.
Technical feasibilityHigh confidence9 / 10
“Prototype running on a live 400 kW feeder since March.”
Quote checked · Matches the descriptor for a working prototype at TRL 6.

Beside the panel’s 46.5, never blended in.

2 more criteria

3 more criteria

What IdeaScore AI Does

The reading a careful reviewer would do, on every application rather than the ten they had time for. People make every decision.

Reviews every application against your rubric
Each criterion is scored against your own descriptors, with the reason in plain words and what would move it up a band.
Shows its evidence
Every score cites the exact words it relied on, and every quote is checked against the application before you see it. Confidence is shown per criterion.
Checks the claims
With Deep Research it tests market size, competitors, patents and named partners against sources you can trust, and tells you what it could not confirm.
Answers your questions
Ask about one application or the whole cohort, such as “Which teams have paying customers?”, and get answers that cite where they come from.
Takes the busywork
Summaries, interview questions, decision-letter and rubric drafts, importing past programmes and setting up a new call from a conversation.
Keeps people in charge
It never decides. Its score sits beside the panel's, never blended in; contact details are never sent to it; you choose whether evaluators see it.

Swift Sorts, Pro Reviews, Deep Decides

One family, three depths of review. Pick a tier for each run, and move the applications that matter to a deeper one.

Swift 4

Fastest

Screen and sort

Screen every application in seconds

A fast first read of your whole intake, so reviewers start with the strongest applications.

recommendations match expert reviewers’, up from about half: 3 in 4
an application: < 30 s

EstimatedEstimated figures, compared with expert reviewers’ scores.

Best for

Large intakes, first sorts and triage

What you get

A score and short reason for every criterion, each with its quote
Every quote checked against the application
A ranking you can sort and shortlist straight away

1credit an application

Step up: Move your shortlist to Pro or Deep before you decide.

Pro 4

Fast

Full evidence-backed review

A full, evidence-backed review of every application

The everyday reviewer: every criterion scored against your descriptors, with the evidence, the confidence and the questions to ask.

of recommendations match expert reviewers’, up from 56%: 84%
of scores within one point of the expert’s, up from 62%: 85%

MeasuredMeasured against expert reviewers’ scores. About 40% faster, and every quote verified.

Best for

Full rounds, feedback to applicants and committee packs

What you get

Everything in Swift, plus the reason and what would move each score up
Confidence per criterion, strengths and risks
Three to five interview questions for each team
Deep Research when you want it

5credits an application

+10 with Deep Research

Step up: Use Deep for the final round, where a few points decide the outcome.

Deep 4.5

Thorough

Researched, reviewed twice

Your toughest calls, researched and reviewed twice

Every claim checked, and every score reviewed a second time against your descriptors and the rest of the round.

recommendations match expert reviewers’: 9 in 10
of scores within one point of the expert’s: 90%

EstimatedEstimated figures, compared with expert reviewers’ scores.

Best for

Shortlists, final rounds, investment and grant decisions

What you get

Everything in Pro, plus Deep Research on every application
A second review that changes a score only with a stated reason, shown beside the original
Scores out of line with the round flagged

30credits an application

Deep Research included

You see the credits before every run. Credits come with every paid plan, in packs, or with IdeaScore AI Plus.

See plans and credits

Model picker

Choose a model in a few words, with its price

Each model shows its name, a few words, its price and what it is best for. Every IdeaScore AI pane fits a phone.Each model in the picker shows its name, a few words on what it does, its price and what it is best for. On a large screen, point at a model to see its details: its speed, its depth and its version. Every IdeaScore AI pane, the picker included, fits a phone screen neatly.

Swift 4: quick first sort, 1 credit
Pro 4: full review with evidence, 5 credits, or 15 with Deep Research
Deep 4.5: researched, two passes, 30 credits

Model picker

Choose a model in a few words, with its price

Each model shows its name, a few words, its price and what it is best for. Every IdeaScore AI pane fits a phone.

Swift 4: quick first sort, 1 credit
Pro 4: full review with evidence, 5 credits, or 15 with Deep Research

Ask Anything About Your Programme

Questions about one application or the whole round, answered with the applications and pages they come from.

IdeaScore AI Chat

Ask about one application or the whole cohort

Answers cite the applications they come from. Deep Research also searches the open web and lists the pages it used; it adds 8 credits on Pro and Deep, shown before you send.IdeaScore AI Chat answers from your programme, citing the applications each answer comes from. Attach an application or a file, rename a chat to keep it, and copy, retry or rate any answer from one quiet row beneath it. Deep Research also searches the web as well as your programme, and lists the pages it used. It adds 8 credits on Pro and Deep, and you see the cost before you send.

Compare, shortlist, find gaps and prepare interview questions
Deep Research lists every page it used
The cost is shown before you send

IdeaScore AI Chat ready to ask across the programme, with suggested questions such as Compare, Shortlist and Find gaps, and the Deep Research switch below the message box.

Questions for This Round

The questions worth asking, written for this round

Three questions about the whole round, each tagged with the part of your rubric it serves. One click opens Chat with it typed in, not sent.On Submissions, Questions for This Round offers three questions about the whole round, each tagged with the part of your rubric it serves, with more under Show more and room to ask your own. A question opens IdeaScore AI Chat with it typed in, not sent, and the teams it cites can be shown in Submissions or saved as a view.

Written from your rubric, your form and the sectors in the round
Typed into Chat for you, never sent on its own
Cited teams can be saved as a view

The Questions for This Round card for the 34 teams in the Semi-final of Meridian Seed Fund 2026: who mentions a patent, which teams got furthest on the least money, and which founders have shipped this kind of product before.

From search

Start from search, with nothing run on its own

Press Ctrl K for suggestions from your programme, or type Chat about and a team. Nothing runs until you send.Press Ctrl K, or Cmd K, and Do with IdeaScore AI offers suggestions built from your programme. Type Chat about and a team to open a new chat with that team attached, or Compare, Shortlist and Find gaps in. Nothing runs until you send.

IdeaScore AI Chat

Ask about one application or the whole cohort

Answers cite the applications they come from. Deep Research also searches the open web and lists the pages it used; it adds 8 credits on Pro and Deep, shown before you send.

Compare, shortlist, find gaps and prepare interview questions
Deep Research lists every page it used

More ways to ask

2 more

Three questions about the whole round, each tagged with the part of your rubric it serves. One click opens Chat with it typed in, not sent.
- Written from your rubric, your form and the sectors in the round
- Typed into Chat for you, never sent on its own
- Cited teams can be saved as a view

Helpers Beside Every Person

IdeaScore AI drafts, explains and suggests in the moment. Nothing is saved, sent or applied until a person confirms.IdeaScore AI drafts, explains and suggests in the moment it is needed, from your programme’s own data. People decide, and nothing is saved, sent, assigned or applied until a person confirms.

A helper at work: pick the awards and the voice, and IdeaScore AI designs a certificate for each, tried on your longest names.

For programme managers and committees

Describe who you’re looking for
A sentence becomes ordinary filters with a live count, before you apply or save a view.
Draft the rationale
A cited rationale from the panel’s scores and notes, saved to the decision note once you confirm.
Draft a funder update
A one-page update with exact cohort figures, cited highlights and risks, and a PDF.
Why do judges disagree?
When two judges are three or more points apart, a cited reading of both sets of notes. Staff only.
What are we missing?
Finds rubric criteria no form question feeds, and offers a question you can undo.
Suggest judges for this team
Three judges with the reason for each; conflicts and capacity are excluded first.
Suggest evaluators
Proposes evaluators for every team in a round, with the reason for each match. One Undo puts it all back.
This Week and Summarise my week
The week’s scoring and letters, worded plainly with a source for each line.
Build a chart with IdeaScore AI
Ask in your own words; IdeaScore counts every number itself, and you pin the chart.

For evaluators and judges

Turn my notes into feedback
Your own notes and the rubric become kind, specific feedback for each criterion, ready to edit.

For mentors

Notes to action items
Session notes become three to five action items with owners and dates, kept or skipped one by one.

For pitch day

Draft team intros
Two-line introductions for each team in running order, for the compère to copy or print.

People decide
A helper never changes a score or decides an outcome.
Free, with fair use
Helpers do not use your credits.
Says what it reads
Every helper’s cost line reads “Free · reads …”, with a How this works note.
Rate and retry
Thumbs and Try again on every result.

Suggest evaluators

A panel for a whole round, with the reason for every match

Evaluators proposed for every team in a round, each with its reason. Accept, swap or skip, then save; one Undo puts it all back.Choose the teams, and IdeaScore AI proposes evaluators for each one, with the reason for the match beside each evaluator’s load. Conflicts and limits are applied first. Accept, swap or skip suggestions one by one, save when you are happy, and a single Undo puts everything back.

What it reads

It tells you exactly what it reads, every time

Every helper says what it reads before you use it, with a How this works note on what it changes.Every helper says in plain words what it reads before you use it, such as “Free · reads judges’ scores and notes on this team”. The line opens a short How this works note on what it reads, what it costs and what it changes.

HydroLoop's dossier on the Criteria tab, with the Why do judges disagree? helper above the score breakdown and the line Free, reads judges' scores and notes on this team.

Two helpers up close

Checks the Claims Against Sources You Can Trust

IdeaScore AI researches the open web, with sources you can check, and prefers official and research sources. Every source is labelled, so you see at a glance how far to trust it.

Official: Startup India and DPIIT, the Ministry of Corporate Affairs, IP India, SEBI, RBI, NITI Aayog and data.gov.in; the World Bank, IMF, OECD and WIPO; patent offices.
Research: Research indexes such as PubMed, arXiv, IEEE and Semantic Scholar.
Press: Reputable business press, for market, funding and competitor news.
Other: Anything else is shown as Other, never as trusted.

Limit answers to official and research sources only, with one switch for your organisation.

Seed Fund 2026 · Cohort III

Research trail

Halcyon Grid · 5 sources used

Official and research only

Size of the feeder-level storage market
data.worldbank.org
Official
No earlier patent on sub-1 MW balancing
ipindia.gov.in
Official
The battery-ageing model holds up
ieeexplore.ieee.org
Research
A competitor raised a seed round
Business news report
Press
Pilot customers named on a blog
Company blog
Other
Named partner: a state utility pilot
Could not be confirmed from public sources
Not confirmed

Why you can trust IdeaScore AI

Evidence, not opinion
Every score quotes the application, and every quote is checked before you see it.
Sources you can judge
Every web source is labelled by trust level; switch to official and research only.
Your data stays yours
Applications are never used to train AI models, and contact details are never sent to IdeaScore AI.
People decide
Its score sits beside the panel’s, never blended in. Re-running never overwrites an earlier result.
No surprises on cost
You see the credits before every run. Failed or timed-out answers are refunded, and questions it declines are free.
Built-in safeguards
Limits on runaway use, protection against misuse, and questions outside IdeaScore’s job are politely declined.

People Decide, and Names Stay with You

People decide. IdeaScore AI does the reading, and applicants are told.

Beside, never instead

Its score sits beside the panel’s, never blended in

It never decides. Evaluators see its notes only if you choose, and applicants are told when it helped. An owner can switch it off for the organisation or one programme.IdeaScore AI never decides. Evaluators don’t see its notes on a team unless your organisation chooses to show them, either after they have submitted their own score or always; blind rounds always hide them. Applicants are told in plain words on the application form, the applicant portal and decision letters when IdeaScore AI helped. An owner can switch IdeaScore AI off for the whole organisation or for a single programme.

Hidden from evaluators by default, or shown after they score, or always
Applicants told on the form, the portal and decision letters
Off for the organisation or a single programme, by an owner

Seed Fund 2026 · Cohort III

Halcyon Grid#0142

CleanTech · TRL 6 · same rubric, out of 50

Human panel

46.5 / 50

3 evaluators, mean

Met the team at the campus clinic. Strong on execution, cautious on the market claim.

IdeaScore AI

44 / 50

6 sources cited · every quote checked

Found two comparable ventures the panel had not seen; neither targets sub-1 MW feeders.

Side by side, never blended

The panel’s score is the score of record.

Names stay with you

Names and contact details never reach IdeaScore AI

Names and contact details are swapped for labels before anything is sent, even inside answers and attachments, and come back only for people allowed to see them.The names and contact details of the people in your programme are never sent to IdeaScore AI, even when they are typed into a free-text answer or an attachment. They are swapped for labels first, and come back only for the people allowed to see them. Every IdeaScore AI window says what it reads.

The IdeaScore AI certificate designer for Meridian Research Park, with a What it reads panel: how many teams, mentors and judges, never who, and a note that applicants' names, answers and scores are never sent.

Beside, never instead

Its score sits beside the panel’s, never blended in

It never decides. Evaluators see its notes only if you choose, and applicants are told when it helped. An owner can switch it off for the organisation or one programme.

Hidden from evaluators by default, or shown after they score, or always
Applicants told on the form, the portal and decision letters

Seed Fund 2026 · Cohort III

Halcyon Grid#0142

CleanTech · TRL 6 · same rubric, out of 50

Human panel

46.5 / 50

3 evaluators, mean

Met the team at the campus clinic. Strong on execution, cautious on the market claim.

IdeaScore AI

44 / 50

6 sources cited · every quote checked

Found two comparable ventures the panel had not seen; neither targets sub-1 MW feeders.

Side by side, never blended

The panel’s score is the score of record.

More on what stays private

1 more

See What It Gave Back

What IdeaScore AI did for your programme, and what it used, in two cards under Usage and Credits.

Time Saved by IdeaScore AI

Reviewer hours given back, as an estimate you can set

Applications read in full, reviewer hours given back and claims checked. You set the minutes a read; the hours stay marked estimated.Time Saved by IdeaScore AI shows the applications read in full, the reviewer hours given back on first reads, and the quotes and claims checked. The hours are always marked estimated: you set how many minutes a first read takes your panel, and a sentence is ready to copy for your report.

The Time Saved by IdeaScore AI card: about 16 hours of reviewer time saved on first reads, 80 applications read in full and every quote checked.

Usage and Credits

Credits, usage and cost per action at a glance

Credits left, usage over time and the cost of each action come first; the rest sits under Where your credits went.Usage and Credits opens on what you check most: the credits left this year, usage over time week by week or day by day, and what each action costs on Swift, Pro and Deep. The rest sits under Where your credits went. You see the cost before every run, failed or timed-out answers are refunded, and credit packs last 12 months.

The Usage and Credits pane of IdeaScore AI: 66,501 of 1,00,000 credits left this year with 5,000 pack credits and Show details, a weekly Usage over time chart of 20,579 credits over the last 12 weeks, and the Cost per action table for Swift, Pro and Deep.

Time Saved by IdeaScore AI

Reviewer hours given back, as an estimate you can set

Applications read in full, reviewer hours given back and claims checked. You set the minutes a read; the hours stay marked estimated.

More on credits and usage

1 more

A Guide That Shows the Change Before It Happens

Ask the Guide; it shows you the change before anything happens.

The IdeaScore AI Guide sits behind Ask in the top bar. It answers how to do anything from the Help Centre, opens the right page and starts tours, and it never reads application data. Now it can make a change for you too.

Ask
“Turn blind review on for the Final.” The Guide knows your programme by name.
Check the preview
Before and after, who it affects, and why you are allowed to make it.
Apply
Nothing happens until you press Apply. The change is recorded in the audit log.
Undo
Changed your mind? Undo is right there.

It only offers what your role allows: people who could not make the change themselves cannot apply it.

The IdeaScore AI Guide on the Rounds page of Meridian Seed Fund 2026 after asking to turn blind review on for the Final: a card shows blind review going from Off to On for 14 evaluators and 26 teams, that you are allowed as an owner, that it is recorded in the audit log and can be undone, with Not now and Apply this change. — Ask the Guide to turn blind review on for a round and it shows the change, who it affects and why you are allowed. Nothing happens until you press Apply.

IdeaScore AI GuideRounds · Meridian Seed Fund 2026 · Cohort III

Rounds

Meridian Seed Fund 2026 · Cohort III

Screening
126 teams · 18 evaluators
Blind on
Semi-final
48 teams · 16 evaluators
Blind on
Final
26 teams · 14 evaluators
Blind off

IdeaScore AI Guide

Help with using IdeaScore

RoundsAbout Meridian Seed Fund 2026 · Cohort III

How can I help?

Ask about Rounds or anything else in IdeaScore. I can open pages and start tours for you.

Suggested for this page

Free. You check every change before it is made.

Owners can stop the Guide offering changes: Settings › Data and privacy › Let the Guide offer changes.

Rounds

Meridian Seed Fund 2026 · Cohort III

Screening
126 teams · 18 evaluators
Blind on
Semi-final
48 teams · 16 evaluators
Blind on
Final
26 teams · 14 evaluators
Blind off

Owners can stop the Guide offering changes: Settings › Data and privacy › Let the Guide offer changes.

Ask the Guide

Everything in IdeaScore AI

Draft letters and rubrics
Decision-letter, rubric and reviewer-summary drafts, never sent automatically.
Fits a phone
Every IdeaScore AI pane fits a phone screen neatly, the model picker included.
Interview questions
Three to five questions for each team, with its strengths and risks.
Set up from a conversation
A new call drafted from a few words, for your kind of programme.
Import past programmes
Bring earlier rounds in so this year starts with history.
Your programme’s documents
Attach your rules, guidelines and briefs; answers keep them in mind and cite them.
Learns your programme
Tuning on Enterprise and Campus, and with IdeaScore AI Plus on Growth and Professional.
Pitch-deck reading
Claims and numbers read from the uploaded deck, slide by slide.
Writing-pattern signal
Shown for your judgement; it never changes a score.
Full scorecard in one click
From any IdeaScore AI result to every judge’s scores and comments.
Build a chart
Describe a chart; IdeaScore counts every number and you pin it to the summary.
Rate and retry
Thumbs and Try again on every helper result and chat answer.
Never overwrites
Re-running keeps every earlier result beside the new one.

Draft letters and rubrics
Decision-letter, rubric and reviewer-summary drafts, never sent automatically.
Fits a phone
Every IdeaScore AI pane fits a phone screen neatly, the model picker included.
Interview questions
Three to five questions for each team, with its strengths and risks.
Set up from a conversation
A new call drafted from a few words, for your kind of programme.
Import past programmes
Bring earlier rounds in so this year starts with history.
Your programme’s documents
Attach your rules, guidelines and briefs; answers keep them in mind and cite them.
Learns your programme
Tuning on Enterprise and Campus, and with IdeaScore AI Plus on Growth and Professional.
Pitch-deck reading
Claims and numbers read from the uploaded deck, slide by slide.
Writing-pattern signal
Shown for your judgement; it never changes a score.
Full scorecard in one click
From any IdeaScore AI result to every judge’s scores and comments.
Build a chart
Describe a chart; IdeaScore counts every number and you pin it to the summary.
Rate and retry
Thumbs and Try again on every helper result and chat answer.
Never overwrites
Re-running keeps every earlier result beside the new one.

What People Ask Before Switching It On