We taught our agent to say "I don't know." It got more accurate.
It once misidentified a candidate confidently, nobody caught it for two days. With ten customers, a wrong guess costs a relationship.
Below a threshold, it now stops and asks a human.
#AIAgents#Founders
Altman expected GPT-4 to shake everything loose in months.
Two years on, he blames the economy's inertia. I read that as trust, not slowness.
A recruiter won't hand a reliable process to something right most of the time. Kritmatta builds for that.
#AI#AIAdoption#Kritmatta
97% agent accuracy sounds impressive, it's the wrong number to chase.
Last month, 97% of what our agents did was right, nobody fixed it. Do 1,000 things at 3% wrong, that's 30 messes to hunt down.
Clever gets the demo. Reliable gets the job done.
#AIAgents#ProductDesign
Your top client posted 11 roles last month. You were briefed on 3.
The other 8 sat public on their careers page, unchecked.
Kritmatta watches career pages continuously, so missing roles surface the moment they go live, not the week before.
#Recruitment#ClientManagement
An 8-person startup had never run an AI agent, just read about them.
That's 99% of SMEs, enterprise AI was written for the 1%. We gave them one agent logging call outcomes.
Two weeks later, nobody had to be reminded to use it. That's nearly everyone.
#AIAgents#SME#Startups
Your AI isn't hallucinating. It's forgetting.
Our agent told a candidate Tuesday, the interview had moved to Wednesday.
Not a bad model, old information. Models are converging anyway.
What matters is what one remembers, not how it thinks.
#AIAgents#Founders#BuildInPublic
An 8-person agency had never run an AI agent, just read about them.
That's 99% of SMEs.
Enterprise AI was written for the 1%.
We gave them one agent logging call outcomes. Two weeks later, nobody had to be reminded to use it.
That's nearly everyone.
We still pitch.
The agent does the convincing.
We put it to work inside someone's own CRM, on their own candidates, before any contract's signed.
Not a demo, the actual thing.
Most software asks you to trust the demo, we'd rather you check yourself.
UK recruitment runs on WhatsApp.
Your CRM has never heard of it.
A placement fell through last year over a go-ahead that lived and died in one group chat.
If your tool doesn't live where the conversation happens, it's not ignored, it's irrelevant.
The day you installed it.
We audited our CRM: 40 records, 11 dead numbers, 6 stale job titles.
A CRM is a Boots receipt, accurate when printed, useless a week later.
Our agent now watches WhatsApp and email and updates itself.
How many of your top 20 would survive a call?
I turned down a paying customer last month.
He wanted an agent for a job that hasn't changed in three years.
That's automation, not an agent. Agents earn their keep where judgement's needed. Using one for his process is a personal trainer at Tesco self-checkout.
#AIAgents
Our agent messaged the same person twice, it was convinced they were two different people.
Cath Hughes: four names across CRM, email, LinkedIn, WhatsApp.
Fixed with bookkeeping, not a smarter model: matching name, email, phone, company.
Count your candidate's versions.
I built an agent so impressive, people clapped in the demo.
It was dead within four days.
The boring one, just logging calls into the CRM, still runs.
Demo instinct kills first agents: you build for the room, not Monday morning.
Pick daily, low-stakes, slightly beneath you.
We promised a candidate feedback by Friday.
We didn't send it.
Commitments have a lifecycle: made, due, done, broken.
Most AI only sees the first.
We built agents that track all four, so a promise stays overdue until kept.
#AIAgents#Recruitment#Accountability
Everyone in sourcing has a take on AI. Fewer will test it in the open.
My co-founder Kalpesh Baxi hosts Vanessa Raath for TEAM Network on where AI pulls its weight and where humans still win.
300+ signed up already!
To sign up now, details here: https://t.co/E9hkKZnMux
Gillette had everything.
A $1 razor took it anyway.
That's Dollar Shave Club. Red oceans have currents big players ignore: an underserved group. Blue oceans reward weak execution; red oceans need real insight.
Which group is your industry ignoring?
#Startups#BlueOceanStrategy
You have a happy place. It's costing you customers.
Mine was coding, always a clear finish line.
A founder I know had WordPress tweaks.
Neither of us thought we were avoiding anything.
What's your happy task? What are you avoiding? Line them up.
Content market fit is a thing. Most creators don't know they're missing it.
Posts don't flop because of the algorithm, they flop because you haven't found what resonates.
To fix: experiment relentlessly, then double down on what works.
Same logic runs Kritmatta's agents.
Most startups fail.
Yours probably will too.
Here's how to make that productive.
Overnight successes on LinkedIn? Look closer. The founder failed multiple times first.
Success is productive failures. Small bets. Fast learning.
Building Kritmatta with that mindset. DM me.
The mistakes I made building Serand made Kritmatta possible.
Serand: wide-ranging, did lots of things poorly.
Kritmatta: fewer things, done better.
Removing features is harder than adding them. More valuable too.
#Startups#ProductStrategy#Kritmatta