Olivia
Online
Simon Deconde
@autthakorn
Senior Software Engineer • Full-stack • SaaS founder | TypeScript, PostgreSQL, React | Product-minded and latency-averse
California
Joined January 2012
15 Following
25 Followers
1.7K Posts
When Ace needs a human decision, like approving an exception, confirming an amount above a limit or reviewing a document it isn't sure about, your team can now respond right from their phone.
Every request shows the context the agent gathered, the data it's relying on and what happens after approval. One tap to approve, reject or send back with a comment, and every decision is logged.
Available now for Automat Workforce customers on iOS and Android.
Hours saved is one number. The backlog tells a better story.
In this illustrative example from one servicing queue, before AI agents, the median open item waited 6.5 days, 42% of items were older than five days and 18% breached their SLA. Sixty days after agents took over the routine work, the median age dropped to 0.8 days, only 3% of items were older than five days and just 1% breached their SLA.
For customers, that's the difference between waiting a week and getting an answer the next day. 📉
Hours saved is one number. The backlog tells a better story.
In this illustrative example from one servicing queue, before AI agents, the median open item waited 6.5 days, 42% of items were older than five days and 18% breached their SLA. Sixty days after agents took over the routine work, the median age dropped to 0.8 days, only 3% of items were older than five days and just 1% breached their SLA.
For customers, that's the difference between waiting a week and getting an answer the next day. 📉
"Hours saved" is the most common automation metric. It's also incomplete. 📊
Hours saved shows cost, but not quality or service. A strong automation program also tracks accuracy (how many completed cases were correct), SLA performance (how many were finished on time), backlog age (how long work waits before someone touches it) and exception rate (how much still needs a person).
These measures show whether automation is improving the business, not just reducing effort. They also tell you where to focus next.
Measure what customers and regulators actually feel.
"Hours saved" is the most common automation metric. It's also incomplete. 📊
Hours saved shows cost, but not quality or service. A strong automation program also tracks accuracy (how many completed cases were correct), SLA performance (how many were finished on time), backlog age (how long work waits before someone touches it) and exception rate (how much still needs a person).
These measures show whether automation is improving the business, not just reducing effort. They also tell you where to focus next.
Measure what customers and regulators actually feel.
Bank operations teams handle a steady stream of account maintenance requests. Each one looks simple, but they add up. 🏦
A change of address, a new beneficiary, an updated phone number, a request to close an account. Each request means verifying the customer, updating the core system and sometimes several other platforms, generating a confirmation and keeping an audit trail.
Automat's agents handle the routine requests across those systems, verify that every update was applied and send anything unusual, like a mismatch in identity details, to an analyst.
Faster service for customers, and fewer small tasks piling up for the team.
8.
Bank operations teams handle a steady stream of account maintenance requests. Each one looks simple, but they add up. 🏦
A change of address, a new beneficiary, an updated phone number, a request to close an account. Each request means verifying the customer, updating the core system and sometimes several other platforms, generating a confirmation and keeping an audit trail.
Automat's agents handle the routine requests across those systems, verify that every update was applied and send anything unusual, like a mismatch in identity details, to an analyst.
Faster service for customers, and fewer small tasks piling up for the team.
8.
How confident should an agent be before it acts on its own? It's a setting, and it's a business decision.
In this illustrative example from one document workflow, a confidence threshold of 0.80 lets agents complete 94% of cases automatically, with 1.6% of errors reaching downstream systems. At 0.90, it's 88% and 0.6%. At 0.95, 79% and 0.2%. At 0.99, only 61% are automatic, but errors drop to 0.05%.
For a marketing report, a lower threshold may be fine. For payments or regulatory data, a higher one is worth the extra reviews.
Set the threshold per workflow, based on what an error really costs.
How confident should an agent be before it acts on its own? It's a setting, and it's a business decision.
In this illustrative example from one document workflow, a confidence threshold of 0.80 lets agents complete 94% of cases automatically, with 1.6% of errors reaching downstream systems. At 0.90, it's 88% and 0.6%. At 0.95, 79% and 0.2%. At 0.99, only 61% are automatic, but errors drop to 0.05%.
For a marketing report, a lower threshold may be fine. For payments or regulatory data, a higher one is worth the extra reviews.
Set the threshold per workflow, based on what an error really costs.
Introducing the Automat Partner Program. 🤝
Consulting firms, systems integrators and automation specialists are increasingly asked by their clients to replace legacy RPA with AI agents. Now they can do it with us.
Partners get training for their teams, access to our forward-deployed engineers on joint projects, early access to new features, co-marketing and dedicated partner support.
If you help enterprises modernize their operations, we'd love to work together. Apply at https://t.co/Upy4QzautJ
Introducing the Automat Partner Program. 🤝
Consulting firms, systems integrators and automation specialists are increasingly asked by their clients to replace legacy RPA with AI agents. Now they can do it with us.
Partners get training for their teams, access to our forward-deployed engineers on joint projects, early access to new features, co-marketing and dedicated partner support.
If you help enterprises modernize their operations, we'd love to work together. Apply at https://t.co/Upy4QzautJ
Operations don't happen in one language, so automation can't either. 🌍
International businesses receive invoices in Spanish, shipping documents in German, insurance forms in French and customer requests in Portuguese. Traditional rule-based bots struggle as soon as the language or layout changes.
Automat's agents read documents and interfaces in multiple languages, extract the right data and enter it into the same systems, with the same validation rules and confidence checks. Anything uncertain goes to a person who speaks the language.
One workflow, many languages, consistent results.
Operations don't happen in one language, so automation can't either. 🌍
International businesses receive invoices in Spanish, shipping documents in German, insurance forms in French and customer requests in Portuguese. Traditional rule-based bots struggle as soon as the language or layout changes.
Automat's agents read documents and interfaces in multiple languages, extract the right data and enter it into the same systems, with the same validation rules and confidence checks. Anything uncertain goes to a person who speaks the language.
One workflow, many languages, consistent results.
Where does an operations analyst's day actually go?
In this illustrative example, before AI agents, 38% of the day goes to navigating portals, 24% to data entry, 14% to checking and reconciling, and only 24% to judgment calls, exceptions and customers. After agents take over the portal work, those shares flip: judgment, exceptions and customers take up 78% of the day.
Same people, same team size. Their time just goes to the work only people can do.
That's the real value of automation. ⏱
Where does an operations analyst's day actually go?
In this illustrative example, before AI agents, 38% of the day goes to navigating portals, 24% to data entry, 14% to checking and reconciling, and only 24% to judgment calls, exceptions and customers. After agents take over the portal work, those shares flip: judgment, exceptions and customers take up 78% of the day.
Same people, same team size. Their time just goes to the work only people can do.
That's the real value of automation. ⏱
The technology is rarely the hardest part of rolling out AI agents. The people side is. 🤝
When agents take over portal work and data entry, jobs on the operations team change. If nobody explains how, people worry, hold back knowledge about exceptions and quietly work around the new system.
What works: involve the team from the first process mapping, be clear about which tasks move to agents and which stay with people, train people to review exceptions and monitor runs, and celebrate the first wins together.
Automation succeeds when the team sees agents as help, not as a replacement.
The technology is rarely the hardest part of rolling out AI agents. The people side is. 🤝
When agents take over portal work and data entry, jobs on the operations team change. If nobody explains how, people worry, hold back knowledge about exceptions and quietly work around the new system.
What works: involve the team from the first process mapping, be clear about which tasks move to agents and which stay with people, train people to review exceptions and monitor runs, and celebrate the first wins together.
Automation succeeds when the team sees agents as help, not as a replacement.
Every new hire triggers a small wave of portal work that most people never see. 🧾
Benefits teams and brokers enroll employees in health, dental and retirement plans across several carrier portals, each with its own forms, rules and deadlines. Life events, like a new baby or a move, mean more updates in every system. Miss an enrollment window, and an employee can be left without coverage.
Automat's agents enter enrollments and changes across carrier and HR portals, check that each one went through and flag anything that needs a person's attention, with a record of every step.
Employees get covered on time. Benefits teams stop retyping the same data five times.
Every new hire triggers a small wave of portal work that most people never see. 🧾
Benefits teams and brokers enroll employees in health, dental and retirement plans across several carrier portals, each with its own forms, rules and deadlines. Life events, like a new baby or a move, mean more updates in every system. Miss an enrollment window, and an employee can be left without coverage.
Automat's agents enter enrollments and changes across carrier and HR portals, check that each one went through and flag anything that needs a person's attention, with a record of every step.
Employees get covered on time. Benefits teams stop retyping the same data five times.
Matching incoming payments to invoices is one of the most manual jobs in finance.
Remittance details arrive in emails, PDFs, customer portals and bank files, often incomplete or combining several invoices into one payment. In this illustrative example, Automat's agents read those details and match 86% of payments to open invoices automatically. Another 10% get a suggested match that a person confirms in one click. Only 4% need to be handled manually.
Faster cash application means a clearer view of cash, fewer customer disputes and a faster month-end close. 💰
Regulatory filings are deadline-driven, repetitive and often spread across dozens of government portals. 🏛
Insurance carriers, lenders and multi-state businesses submit licensing renewals, rate filings, tax forms and compliance reports to different state agencies, each with its own website, format and schedule. Missing a deadline can mean fines or lost licenses.
Automat's agents track deadlines, fill out and submit filings across portals, download confirmations and keep a complete audit trail for every submission.
Compliance teams stay in control. The agents handle the forms.
Matching incoming payments to invoices is one of the most manual jobs in finance.
Remittance details arrive in emails, PDFs, customer portals and bank files, often incomplete or combining several invoices into one payment. In this illustrative example, Automat's agents read those details and match 86% of payments to open invoices automatically. Another 10% get a suggested match that a person confirms in one click. Only 4% need to be handled manually.
Faster cash application means a clearer view of cash, fewer customer disputes and a faster month-end close. 💰
Matching incoming payments to invoices is one of the most manual jobs in finance.
Remittance details arrive in emails, PDFs, customer portals and bank files, often incomplete or combining several invoices into one payment. In this illustrative example, Automat's agents read those details and match 86% of payments to open invoices automatically. Another 10% get a suggested match that a person confirms in one click. Only 4% need to be handled manually.
Faster cash application means a clearer view of cash, fewer customer disputes and a faster month-end close. 💰
Automat is opening an office in London. 🇬🇧
Banks, insurers and financial services firms across the UK and Europe run on the same mix of legacy systems, portals and manual work that we automate every day, often with strict requirements on data residency and oversight.
Our London team will work directly with customers across the region, with forward-deployed engineers on the ground and the same platform, security standards and support.
We're hiring engineers and customer-facing roles in London. Open roles at https://t.co/f8ZqTxCpq1
Automat is opening an office in London. 🇬🇧
Banks, insurers and financial services firms across the UK and Europe run on the same mix of legacy systems, portals and manual work that we automate every day, often with strict requirements on data residency and oversight.
Our London team will work directly with customers across the region, with forward-deployed engineers on the ground and the same platform, security standards and support.
We're hiring engineers and customer-facing roles in London. Open roles at https://t.co/f8ZqTxCpq1
When an AI agent makes a decision, your team should be able to see why. 🔍
Every Automat run can include a plain-language explanation of the key decisions: why a document was classified a certain way, why a case was sent for review, which rule triggered an exception and which data the agent relied on.
For operations teams, that makes reviewing exceptions faster. For compliance and audit, it turns automation from a black box into a clear record.
Trust in automation comes from understanding it.
When an AI agent makes a decision, your team should be able to see why. 🔍
Every Automat run can include a plain-language explanation of the key decisions: why a document was classified a certain way, why a case was sent for review, which rule triggered an exception and which data the agent relied on.
For operations teams, that makes reviewing exceptions faster. For compliance and audit, it turns automation from a black box into a clear record.
Trust in automation comes from understanding it.
Here's the agenda for the Automat Agents Summit on November 12 in San Francisco.
We'll open at 9:00 with a keynote on what really changes when you move from legacy RPA to AI agents. At 10:00, a live demo of Ace working across real enterprise portals. At 11:00, a panel with operations leaders from banking, insurance and mortgage. After lunch, a hands-on workshop with our forward-deployed engineers on scoping your first agent workflow, followed by a session on security, compliance and audit. We'll close with a roadmap preview and networking.
Request an invite at https://t.co/Upy4Qzb2jh 🗓
Here's the agenda for the Automat Agents Summit on November 12 in San Francisco.
We'll open at 9:00 with a keynote on what really changes when you move from legacy RPA to AI agents. At 10:00, a live demo of Ace working across real enterprise portals. At 11:00, a panel with operations leaders from banking, insurance and mortgage. After lunch, a hands-on workshop with our forward-deployed engineers on scoping your first agent workflow, followed by a session on security, compliance and audit. We'll close with a roadmap preview and networking.
Request an invite at https://t.co/Upy4Qzb2jh 🗓
Business rules change all the time: a new approval limit, a new document requirement, a new portal. How fast can your automation keep up? ⚡
With legacy RPA, a small rule change often means a ticket, a developer and a new release weeks later. With Automat, many changes are updated in hours, because agents work from instructions and rules, not hard-coded scripts, and our forward-deployed engineers stay close to your team.
Every change is tested, versioned and logged before it goes live.
Automation should move at the speed of your business, not slow it down.
Business rules change all the time: a new approval limit, a new document requirement, a new portal. How fast can your automation keep up? ⚡
With legacy RPA, a small rule change often means a ticket, a developer and a new release weeks later. With Automat, many changes are updated in hours, because agents work from instructions and rules, not hard-coded scripts, and our forward-deployed engineers stay close to your team.
Every change is tested, versioned and logged before it goes live.
Automation should move at the speed of your business, not slow it down.
In e-commerce, most orders flow through without a problem. The ones that don't can eat an entire team's day. 📦
An address that doesn't validate, a payment flagged for review, an item out of stock at the chosen warehouse, a customer asking to change their order after it's been sent to fulfillment. Each exception means checking several systems and portals before anyone can act.
Automat's agents investigate each exception, gather the information from every system, resolve the routine cases automatically and send the rest to a person with everything they need to decide.
Customers get their orders faster. Teams stop firefighting.
In e-commerce, most orders flow through without a problem. The ones that don't can eat an entire team's day. 📦
An address that doesn't validate, a payment flagged for review, an item out of stock at the chosen warehouse, a customer asking to change their order after it's been sent to fulfillment. Each exception means checking several systems and portals before anyone can act.
Automat's agents investigate each exception, gather the information from every system, resolve the routine cases automatically and send the rest to a person with everything they need to decide.
Customers get their orders faster. Teams stop firefighting.
When does automation pay for itself?
In this illustrative example, one workflow costs $40K to build and deploy, frees up about $15K a month in team capacity and costs $3K a month to run. Break-even comes around month 3.3. By month 12, the cumulative value reaches about $104K.
The exact numbers vary by process. The shape usually doesn't: a short investment period, then steady returns every month after.
That's why we start every engagement with the business case, not the technology. 📈
When does automation pay for itself?
In this illustrative example, one workflow costs $40K to build and deploy, frees up about $15K a month in team capacity and costs $3K a month to run. Break-even comes around month 3.3. By month 12, the cumulative value reaches about $104K.
The exact numbers vary by process. The shape usually doesn't: a short investment period, then steady returns every month after.
That's why we start every engagement with the business case, not the technology. 📈
A single real estate transaction can require searching records across several county portals, each with its own interface. 🏠
Title teams look up deeds, liens, judgments and tax records, download documents and piece together the chain of title, often on websites that were built decades ago and have no API.
Automat's agents search those portals, download the relevant documents, extract the key details and assemble them into a structured report for the title examiner to review.
The examiner still makes the judgment. The agent does the searching.
A single real estate transaction can require searching records across several county portals, each with its own interface. 🏠
Title teams look up deeds, liens, judgments and tax records, download documents and piece together the chain of title, often on websites that were built decades ago and have no API.
Automat's agents search those portals, download the relevant documents, extract the key details and assemble them into a structured report for the title examiner to review.
The examiner still makes the judgment. The agent does the searching.
Mortgage and loan servicing teams spend hours on requests that follow the same steps every time. 🏦
A payoff statement request means pulling the balance, calculating per-diem interest, checking fees and escrow, generating the document and sending it before the deadline. Escrow analyses, insurance updates and borrower requests work the same way, across several systems and portals.
Automat's agents handle those steps end to end, with calculations checked against the servicing system and exceptions routed to a specialist.
Faster answers for borrowers, fewer errors and less overtime at month-end.
Mortgage and loan servicing teams spend hours on requests that follow the same steps every time. 🏦
A payoff statement request means pulling the balance, calculating per-diem interest, checking fees and escrow, generating the document and sending it before the deadline. Escrow analyses, insurance updates and borrower requests work the same way, across several systems and portals.
Automat's agents handle those steps end to end, with calculations checked against the servicing system and exceptions routed to a specialist.
Faster answers for borrowers, fewer errors and less overtime at month-end.
Trends for you
Most Popular Users
Elon Musk
@elonmusk
241.7M followers
Barack Obama
@barackobama
119M followers
Cristiano Ronaldo
@cristiano
114.4M followers
Donald J. Trump
@realdonaldtrump
111.9M followers
Narendra Modi
@narendramodi
107.2M followers
Rihanna
@rihanna
98.7M followers
NASA
@nasa
92.4M followers
Justin Bieber
@justinbieber
91.8M followers
KATY PERRY
@katyperry
90M followers
Taylor Swift
@taylorswift13
84M followers
Lady Gaga
@ladygaga
75.4M followers
Virat Kohli
@imvkohli
73.4M followers
Kim Kardashian
@kimkardashian
70.9M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
66.4M followers
Bill Gates
@billgates
65.2M followers
Selena Gomez
@selenagomez
63.1M followers
The Ellen Show
@theellenshow
62.3M followers
CNN
@cnn
61.8M followers
X
@x
60.7M followers