Adopt AI CEO Unveils Framework to Replace Manual Finance Workflows
Ai4
Enterprise finance operations require autonomous execution agents rather than rigid task automation to eliminate manual data coordination and accelerate month-end close cycles.
## Background
This video was published on YouTube by the channel Ai4 on August 15, 2026. Featuring Deepak Anchala, Co-founder & CEO of Adopt AI, the presentation explores how finance operations can move beyond traditional ERP investments and manual workflows to achieve measurable value through autonomous execution agents.
## TL;DR
Deepak Anchala addresses the persistent issue of manual, fragmented finance workflows that slow down month-end closing and FP&A processes despite heavy software investments. He outlines the core pillars of autonomous execution—Execute, Verify, and Govern—showing how AI agents can handle context-gathering, verification, and end-to-end task coordination. The presentation details high-value finance use cases like close management, reconciliations, and AP/AR automation, concluding with a practical evaluation framework based on Rule Clarity and Handoff Density to determine where to deploy AI execution.
## Atmosphere & style
The session carries a professional, authoritative, and consultative tone typical of an enterprise tech keynote. Deepak presents complex workflow issues with structured clarity, blending pragmatic diagnostic insights with reassuring guidance on governance and control. The presentation feels focused, practical, and solution-driven, maintaining a clear educational arc aimed directly at finance leaders.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=4s) Major enterprise investments in ERPs, automation, and analytics have failed to eliminate manual, fragmented finance operations.
- [00:04](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=4s) The primary bottleneck in finance workflows is the heavy prep work and manual cross-system coordination, not generating reports.
- [00:04](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=4s) Industry statistics show that 94% of organizations still rely heavily on Excel spreadsheets to bridge gaps between core enterprise systems.
- [00:04](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=4s) Traditional RPA and Business Process Automation (BPA) operate as thin task-level layers that fail to execute end-to-end processes autonomously.
- [04:31](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=271s) Autonomous execution relies on three core pillars: Execute (end-to-end workflow execution), Verify (accuracy and business rules enforcement), and Govern (auditability and C-suite controls).
- [04:31](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=271s) Autonomous AI agents can perform deterministic tasks alongside non-deterministic coordination in a tightly controlled and cost-effective manner.
- [04:31](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=271s) Demonstration shows agents automatically extracting bank statement and ERP data to generate complete general ledgers and trial balance spreadsheets.
- [08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) Deploying autonomous agents in the financial close replaces sequential human checklists and can reduce multi-day close cycles to hours.
- [08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) High-value use cases for autonomous execution include reconciliations, accounts payable (AP), accounts receivable (AR), and FP&A data prep.
- [08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) A practical evaluation framework measures Handoff Density, Rule Clarity, Process Volume/Frequency, and Cost of Error to determine automation strategies.
- [08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) Workflows with High Handoff Density and High Rule Clarity (e.g., AP matching, close) are prime candidates for full autonomous execution.
- [08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) Low handoff density with low rule clarity workflows offer virtually no ROI and should not be automated.
- [13:48](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=828s) Adopt AI uses patented technology to automatically discover browser APIs and convert them into usable endpoints for seamless integration.
- [13:48](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=828s) The platform functions as a multi-agent, multi-human collaborative harness allowing human preparers and reviewers to retain strict governance.
## Detailed walkthrough
Here is a detailed, faithful walkthrough based on the transcript provided:
[00:04](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=4s) The Financial Execution Paradox and Diagnosing the Problem
Deepak Anchala, co-founder and CEO of Adopt AI, opens his presentation by asking the audience of finance leaders how many of them have invested heavily in ERPs, automation, analytics, or AI tools. Despite decades of massive enterprise investments, everyday finance processes remain stubbornly manual. Task-level reconciliations still require heavy human intervention, and the month-end close continues to crawl along, heavily constrained by whichever individual finishes their close checklist last. As a result, running board reports and closing the books still takes many days. To address why organizations are stuck in this state despite the promises of AI, Deepak outlines four key takeaways for the session: diagnosing the structural flaw in finance workflows, defining the paradigm shift from traditional automation to autonomous execution, identifying where actual ROI exists for finance leaders, and providing a practical framework to evaluate which enterprise processes to transform first.
When diagnosing the root cause, Deepak notes that while software solutions and spreadsheets proliferate, the core bottleneck is never the actual reporting layer. Generating a report itself is simple; the actual "time suck" is the immense prep work and "grunt work" required beforehand. This involves manually pulling data across disparate systems—such as CRMs, ERPs, and record-keeping software—and coordinating directly with people to resolve inconsistencies or answer routine questions. Ultimately, this is an execution and coordination problem rather than a lack of data or personnel. The necessary facts reside in emails, ERPs, and bank systems, but tying those threads together, handling exceptions, and getting questions answered currently forces teams to live inside Excel spreadsheets.
- Industry research shows that it takes an average of six days for organizations to close their books.
- Only 18% of organizations manage to complete their month-end close in under three days.
- A staggering 94% of organizations still rely heavily on Excel spreadsheets to bridge gaps between their core enterprise systems.
- Existing tools like Business Process Automation (BPA) and Robotic Process Automation (RPA) operate purely as a thin layer on top of software to automate isolated, single-step tasks (e.g., clicking a button to execute a short script), failing to execute end-to-end processes autonomously or deliver genuine ROI.
---
[04:31](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=271s) Pillars of Autonomous Execution and Governance
To move beyond rigid task automation, Deepak introduces the three core pillars of autonomous execution: **Execute**, **Verify**, and **Govern**. Execution involves connecting autonomous agents directly to underlying data sources so they can gather context, query systems, apply business logic, and execute workflows end-to-end without constant human intervention. Because finance demands high precision, a robust verification layer must constantly double-check that source data remains accurate, transformation logic is correct, and company rules are strictly applied. Lastly, governance provides enterprise oversight through detailed audit logs, random sampling, and controls that let C-suite executives manage agents just as they would human employees. Deepak emphasizes to skeptical finance leaders that "autonomous does not mean unpredictable," reassuring them that agents can run deterministic tasks (such as calling bank statement APIs or pulling ERP records) alongside non-deterministic task coordination in a tightly controlled, cost-effective manner to limit token usage.
Deepak then shares a video preview demonstrating an autonomous workflow in action. In the video, an agent orchestrates a full workflow by extracting context from bank statements and ERP systems to build comprehensive general ledgers and trial balances on the fly. Rather than forcing staff to manually construct complex multi-tab Excel workbooks, the agent draws from past Excel templates, existing business logic, and historical workflows to automatically generate identical spreadsheets or update systems of record directly. From there, the output feeds into automated tax form generation, flux analysis, and broader FP&A planning.
---
[08:47](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=527s) High-Value Finance Use Cases and the Transformation Framework
Deepak details specific operational areas where autonomous agents drive immediate, measurable value across finance operations:
- **Financial Close:** Replaces sequential, human-dependent close checklists with autonomous agents that work continuously overnight or through weekends, reducing multi-day close cycles down to mere hours.
- **Reconciliations:** Automatically ingests invoice emails, bank statements, and ERP line items, applying predefined matching rules to resolve item-level discrepancies.
- **Accounts Payable (AP):** Automates invoice OCR extraction, validates contract rules, flags overspending, and routes high-threshold exceptions straight into enterprise platforms like Workday.
- **Accounts Receivable (AR):** Manages cash flow by autonomously tracking customer payments, issuing billing notifications, and sending ongoing email follow-ups.
- **FP&A:** Eliminates manual data prep work so teams can focus directly on forecasting, budgeting, and cash flow analysis.
To help leaders decide where to deploy autonomous execution, Deepak presents a four-element evaluation framework consisting of **Handoff Density** (the level of cross-team coordination needed), **Rule Clarity** (how clearly defined the underlying business rules are), **Process Volume/Frequency**, and **Cost of Error**. Mapping these on a two-by-two matrix of Rule Clarity versus Handoff Density yields four distinct operational strategies:
1. **Transform with Autonomous Execution (High Handoff Density / High Rule Clarity):** Workflows with clear rules but heavy manual coordination—such as AP matching, reconciliations, and financial close—are prime candidates for full autonomous execution with a human reviewer at the end.
2. **Standard Automation / RPA (Low Handoff Density / High Rule Clarity):** Highly deterministic, low-coordination tasks should rely on standard RPA tools rather than complex AI agents.
3. **AI Augmentation (High Handoff Density / Low Rule Clarity):** Processes requiring significant human judgment and possessing lower rule clarity (such as FP&A analysis or complex exception handling) should use AI only to assist human decision-makers, not replace them.
4. **Do Not Automate (Low Handoff Density / Low Rule Clarity):** Low-volume tasks with ambiguous rules offer virtually no ROI and should not be automated.
---
[13:48](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=828s) The Adopt AI Platform and Closing Remarks
Contrasting legacy RPA bots—which break as soon as a user interface changes or a workflow requires non-deterministic execution—Deepak explains that true autonomous execution manages end-to-end workflows from start to finish while looping humans in for review. He introduces Adopt AI's platform, which features over 300 pre-built integrations across the finance ecosystem alongside patented technology that automatically discovers browser APIs and converts them into workable API endpoints. Operating as a multi-agent, multi-human collaborative environment, the harness allows preparers, reviewers, and administrators to govern AI agents effectively.
Deepak closes his talk by inviting attendees to visit Adopt AI at **Booth 729** for a customized demo and swag, or to scan the on-screen QR code to schedule an online demonstration.
## Notable quotes
> [04:31](https://www.youtube.com/watch?v=0a1OdvsNA1o&t=271s) autonomous does not mean unpredictable — Deepak Anchala
## Video description
Deepak Anchala - Co-founder & CEO, Adopt AI
What if your next finance transformation didn't require replacing your ERP?
Finance has spent decades investing in ERP, automation, and analytics, yet many of its most critical processes remain manual, fragmented, and dependent on human coordination.
Join leaders from Adopt AI to explore where autonomous agents are creating measurable value across finance operations and learn a practical framework for identifying the workflows that should be transformed first.
See how leading finance teams are moving beyond automation to autonomous execution.
Adopt AI CEO公布替代人工财务工作流的框架
企业财务运营需要自主执行智能体而非僵化的任务自动化，以消除人工数据协调并加速月末结账周期。
Deepak Anchala针对即便投入大量软件资源，人工且碎片化的财务工作流仍会拖慢月末结账和FP&A流程这一长期痛点进行了探讨。他概述了自主执行的三大核心支柱——执行（Execute）、验证（Verify）与治理（Govern），展示了AI智能体如何胜任上下文收集、核验以及端到端任务协调。该演讲详细阐述了结账管理、对账以及应收/应付账款（AP/AR）自动化等高价值财务用例，并最后提出了一个基于规则明确度（Rule Clarity）和交接密集度（Handoff Density）的实用评估框架，用以确定应在何处部署AI执行。
Deepak Anchala 指出，尽管企业在 ERP 与自动化上投入巨大，但 94% 的组织仍依赖 Excel 填补系统鸿沟；财务流程的核心瓶颈在于繁重的数据预处理与跨系统人工协调，而非报表生成本身。 (0:04)
区别于仅能处理单步孤立任务的传统 RPA，自主执行（Autonomous Execution）依托“执行、验证与治理”三大支柱，使 AI 智能体在严格审计与风控下端到端完成复杂业务流。 (4:31)
在财务结账、对账和应付账款（AP）等高协调密度且规则清晰的场景中，自主智能体可实现连续跨夜运行，将长达数天的结账周期缩短至数小时。 (8:47)
Adopt AI 推出具备 300 多个财务生态预构建集成与专利 API 发现技术的平台，构建多智能体与人类协同环境，使管理层能够像管理员工一样规范和审查智能体工作。 (13:48)
Reading summary
Adopt AI co-founder Deepak Anchala argues that finance operations remain bottlenecked by manual cross-system data preparation and coordination, necessitating a shift from single-step robotic automation to governed autonomous execution.
The Cross-System Coordination Bottleneck in Finance (0:04)
Despite substantial investments in ERP systems and business intelligence tools, routine finance workflows remain heavily reliant on manual intervention. Anchala points out that generating reports is straightforward; the primary operational bottleneck is the extensive data gathering, exception handling, and human coordination required beforehand across CRMs, ERPs, and banking systems. Because existing tools fail to coordinate these disparate systems, finance teams continue to bridge operational gaps through spreadsheets.
To illustrate the problem, Anchala cites industry research showing that closing the books takes organizations an average of six days, with only 18% completing the process in fewer than three days. Furthermore, 94% of organizations still rely on Excel spreadsheets to bridge enterprise systems. Anchala asserts that legacy Business Process Automation (BPA) and Robotic Process Automation (RPA) merely add a fragile layer over software to execute isolated clicks or scripts rather than handling end-to-end workflows.
Pillars of Autonomous Execution and Governance (4:31)
To replace fragmented task automation, Anchala introduces three foundational operational pillars: Execute, Verify, and Govern. Under this framework, autonomous agents connect directly to data repositories to retrieve context, query systems, and apply business logic end-to-end. To address precision requirements in accounting, a dedicated verification layer checks source data, business logic, and internal policies, while governance mechanisms provide audit trails and sampling controls comparable to human management.
Addressing enterprise skepticism regarding AI reliability, Anchala maintains that autonomous execution does not mean unpredictable behavior. The system pairs deterministic tasks—such as pulling ERP records or calling bank APIs—with controlled non-deterministic orchestration designed to manage token consumption and prevent workflow deviations.
A Framework Based on Handoffs and Rules (8:47)
Anchala outlines an evaluation matrix based on Rule Clarity, Handoff Density (cross-team coordination requirements), Process Volume, and Cost of Error. Workflows characterized by high rule clarity and high handoff density represent prime targets for autonomous execution, supported by a final human sign-off. These include month-end financial closes, transaction matching in reconciliations, accounts payable invoice extraction and validation in platforms like Workday, and accounts receivable payment tracking.
Conversely, workflows outside this quadrant require alternate strategies. High-clarity tasks with minimal handoffs are best served by traditional RPA. Low-clarity, judgment-intensive workflows—such as qualitative FP&A analysis or complex exception handling—warrant AI augmentation rather than autonomous agency. Tasks that have both low rule clarity and low handoff density yield negligible return on investment and should not be automated.
Platform Architecture and Human-in-the-Loop Oversight (13:48)
Traditional RPA bots are vulnerable to workflow failure whenever user interfaces shift. Anchala contrasts this brittleness with an architecture built around more than 300 pre-built enterprise integrations and automated browser API discovery. This setup establishes a collaborative framework that links autonomous agents with human preparers, reviewers, and administrators to ensure that autonomous operations remain subject to direct human review.
Anchala concludes that autonomous execution can compress multi-day close cycles into hours, but emphasizes that successful enterprise adoption depends on confining autonomous workflows to rule-bound processes while preserving rigorous verification and human-in-the-loop governance.
[4:31] Deepak Anchala: autonomous does not mean unpredictable
Original transcript
[0:04] Hey everyone. I'm founder CEO of Adopt AI and
I'm here to talk about finance automation.
[0:10] Actually, autonomous agents in finance moving
away from automation to autonomous execution.
[0:15] So I have a simple question for all of you.
[0:17] I'm assuming a lot of this room is finance leaders
here. How many of you have invested in an ERP?
[0:25] Automation analytics AI probably, right? A
lot of you have. But still even even with all
[0:32] these investments, today's finance process is
still manual. Like reconciliations are manual.
[0:37] Month end close depends on the last person doing
their close checklist, right? To move things
[0:42] forward. It still takes many days to close
and and you know, do your board reporting.
[0:47] So why? And why are why we still living in this
world when we hear everything about AI and and
[0:53] what autonomous agents can do? So I'm going
to answer that question and I'm also going
[0:57] to give you guys a framework of how to think
about autonomous execution in your enterprise.
[1:04] So what you'll walk away with today are four
things. One is a clear diagnosis of the problem
[1:10] statement. Where exactly is the problem
in finance and especially when it comes to
[1:14] closing reconciliations bookkeeping. What
is the shift that we are going to see
[1:21] from automation to autonomous agents? Where is
where is value? Because as CFOs, you must be
[1:26] questioning when whenever you talk to every other
team, you question what's the ROI, right? Like is
[1:30] there a cost benefit? Is there revenue addition?
But when you think of your own process too
[1:35] and running autonomous agents, you that's the
first thing in your mind. Where is the ROI?
[1:39] Many of you have small teams. So where is ROI
and how do how should you see look at value here?
[1:44] And finally, we'll come come out try to
come up with a framework that you can
[1:48] apply to your organization to see
what processes you can transform.
[1:54] So let's diagnose the problem a little bit, right?
Um, there's been decades of investment in the
[1:59] space. ERPs have existed forever. Um, analytics
tools, you probably use record keeping software,
[2:06] um, financial close software. There's so much
software around and there's people around.
[2:11] But, even today it still takes on an average
6 days to close books. This is a statistic um,
[2:18] that is out there. Um, research shows that 18% of
the organizations take less than 3 days to close.
[2:24] And 94% of the organizations still rely on Excel
spreadsheets in spite of all the systems here,
[2:30] right? So, there is definitely something
missing in this in this whole system.
[2:36] Um, if you um, look at where the bottleneck is,
the bottleneck is never the reporting layer,
[2:42] right? The reports are you you can
generate them. It's all the prep work,
[2:45] the grunt work that you do to get to reporting.
Which is manual coordination. You have to go
[2:50] when you want to reconcile expenses,
invoices, you have to talk to people,
[2:54] you have to get stuff done. Uh, you have to
pull reports from different systems, CRM, ERP,
[2:59] and reconcile. There's a lot of grunt work
and prep work that is really a time suck here.
[3:05] Um, the the basic problem is it's it's a it's not
a it's a coordination and execution problem. It's
[3:11] not a data problem. It's not a people problem. The
data is there. The this the true source of data
[3:17] is is also there in all these applications from
email to ERPs, to all these systems. And you also
[3:22] have the people. The problem is how do you piece
all these things together, make sense out of it,
[3:27] resolve exceptions automatically, go to people
that need to answer questions and get those
[3:33] questions answered, and and how do you automate
the whole process, right? So, it's it's a major
[3:37] coordination problem. Uh, which is why we live
in Excel for a large part, right? Excel helps us
[3:43] get everything together. We run our own
formulas, our own logic, and then um,
[3:47] and and then get it to a shape there, you know,
you're probably good with the numbers, right?
[3:54] And if you look at you know, automation has
always existed, right? We had BPA, then RPA,
[4:00] all kinds of automation still exist today.
But if you look at what automation has done,
[4:04] automation has helped you automate specific
tasks in the ERP or in different systems,
[4:08] right? When you click on a button, you do these
five things, I'm running a script to automate it.
[4:13] But it hasn't really taken the end-to-end process
and executed it more autonomously, which today
[4:19] you can do. And we're going to talk about how you
do that. Um but automation was just in my mind
[4:24] just another layer on top of software. It really
didn't give you the ROI that you're looking for.
[4:31] So, what is autonomous execution and how is
it different, right? So, there are three main
[4:35] aspects for autonomous execution. Execute, verify,
and govern. Execute is you connect the autonomous
[4:42] agents to all the systems that you need to connect
to all the source of data, so the agents can
[4:47] autonomously, without human intervention,
go and extract information, ask questions,
[4:53] get stuff done, you know, they they work on
logic and they execute and deliver outputs.
[4:58] Two, verification. I think from finance,
you want 100% or close to 100% accuracy.
[5:03] So, you need at every step of the automation, you
need a layer that verifies if your rules are being
[5:08] applied, if the source data is accurate or not,
transformations are accurate. And and finally,
[5:13] governance, right? Whenever you're doing any
aspect like this, you want audit logs. You want
[5:18] to go in make sure everything is working well.
You want random sampling. You want the ability to
[5:22] control the agents, just like you control your
team members. Uh and you want trust that, you
[5:27] know, these can execute autonomously in a manner
that, you know, that is right for the company.
[5:32] So, those are the fundamental layers in autonomous
execution, even more so for C-suite and finance
[5:38] folks, because accuracy is so important here. And,
you know, you you mess up things, then the whole,
[5:43] you know, company goes for for You don't want it
the wrong numbers. Um I also want to men- mention
[5:49] that autonomous does not mean unpredictable. Like
a lot of people get scared when they hear the
[5:53] moment autonomous exe- execution. You can still
operate both deterministic and non-deterministic
[5:59] processes in a very governed manner. If you
have the right set of rules, you apply these
[6:04] rules and checks and balances, and you you have
like specific processes that are deterministic,
[6:10] uh just like, you know, like for example, like
you're pulling bank statements from bank account,
[6:14] you're pulling transactions from an ERP, you're
for flux analysis you're getting CRN's data.
[6:18] All of these are API calls, right? Most of them.
You can't go wrong on those because you're calling
[6:23] statements API, you're calling uh But then, which
API do you call? How do you call them? That's the
[6:28] non-deterministic portion that, you know, humans
do today. And that's something autonomous agents
[6:33] can also be trained to do when they're automating
the grunt work. So, you still have to operate it
[6:38] in a very rule-based deterministic manner
so you don't consume too many tokens,
[6:42] right? You want to save on token cost. We'll
also have the non-determinism made you want to,
[6:46] and finally, make make it govern um make make
it auditable. I'm going to show you a quick
[6:53] demo of a process here which shows you how
you can move from, you know, getting bank
[7:00] statements to talking to ERPs, creating balance
sheets, trial balances, you know, on the fly,
[7:06] and what can you do here? It's not a full demo.
Uh I'll show you how you can get a full demo, but
[7:10] it's just a preview of, you know, what autonomous
agents can do here. It's um I'm going to start
[7:15] the video. Okay. This is just a workflow. It's
a simple workflow. And um this is basically um
[7:23] you know, an agent coordinating the work where
it's gathering the information from required
[7:28] systems, understanding the context, and then,
instead of waiting for a person to do this, it's
[7:35] going and doing this. For example, here it created
an entire um I think this is um a general ledger
[7:42] and a trial balance where it creates by looking at
bank statements, getting information from the ERP.
[7:48] The agents are creating all these documents. We
create like complex Excel spreadsheets, balancing
[7:54] books with multiple tabs. And the reason they're
able to do that is because they're trained on
[7:59] your business logic, on your prior Excel files, on
how your team has done it, and then recreate the
[8:03] same Excel. If you don't want Excel, they can go
and do this on a system of record that you have.
[8:09] Um and as you can see here, you know,
the the information is all all here. Um
[8:15] and then you can go and start like even doing
your I think it's still doing the balance sheet.
[8:21] Uh it's generating all the statements. Post that,
you can also use it for tax filing, right? You can
[8:27] create tax forms. You can do flux analysis post
that. Um so there's a lot of FP&A uh planning and
[8:33] reporting. Um but all of these documents can
be created automatically using the same rules
[8:39] um that a human would, right? It understands your
Excel formats and recreates that. So, moving on.
[8:47] So, where does autonomous execution in finance
create value? Uh these are These are a few
[8:52] sample processes. So, for example, financial
flows, we talked about a close being 6 days.
[8:58] You could potentially bring it down to hours,
right? If you have the right processes in place,
[9:02] the moment, you know, month ends, you can go and
grab all the information from different systems.
[9:07] Uh generally, financial close is restricted the
the time the it takes long because the slowest
[9:13] person on the team has to finish their job in
the close checklist before the other person can
[9:17] take take that and move move on, right? But here,
the agents are autonomously doing it. You know,
[9:21] let's say your financial close is on Friday, they
work on the weekend, they're working overnight
[9:25] when humans are sleeping, and they're generating
everything that you need for for your close.
[9:31] Reconciliations today is very manual. Uh that's
another process where, you know, you can pull,
[9:36] for example, email invoices from email
or information from bank accounts, ERPs,
[9:41] start reconciling item by item with predefined
set of rules for these agents. Accounts payable
[9:48] is another place where we see a lot of people
deploy agents. Uh you obviously have rules
[9:53] when you're paying vendors. You You don't want to
overshoot them. You want to check contracts. Yeah,
[9:58] you want to process invoices the right way.
Sometimes you have to OCR extract information from
[10:02] those invoices, right? From your AP ID, flag them
and see and and then in a lot of cases you have to
[10:08] route those invoices for exception handling,
right? If they're above a certain threshold.
[10:12] Agents can do all of that and they can interact
with Workday or whatever your system is to kind
[10:17] of, you know, trigger the right right workflows
there. Same with accounts receivable. Accounts
[10:22] receivable, you want to manage your cash more
conservatively. You want to make sure you have
[10:26] cash in the bank in in some instances. Make sure
Sorry. You want You want to get cash, you know,
[10:33] from your customers when when the time is right
and you You want to bill them the right way.
[10:38] So, you can you can chase customers, right?
There's a lot of coordination where you're
[10:41] continuously asking customers to pay. Agents
can do all of that coordination. They can send
[10:45] emails out. They can report back. Uh and and
so you can just govern and manage those agents.
[10:51] And finally, FP&A where a lot of time today is
also spent in prep work. Uh all of this would
[10:56] happen after your financial flows, forecasting,
budgeting, cash flow analysis for the team.
[11:04] So, when Let's look at the framework. When When
you're looking at all of these processes, like
[11:08] how do you determine where do you start and what
do you What do you do autonomously using agents?
[11:14] So, there's two slides here. The first one
talks about the four elements and then we'll
[11:18] place the elements on a two by two matrix,
like, you know, every consulting firm does.
[11:23] Um so, the first one is handoff density,
right? Does the process that you have require
[11:28] a lot of coordination between teams, between
people? If yes, then that's a someplace where
[11:34] you don't want, you know, to spend human
time on it, right? You want an automated
[11:38] process doing all of that coordination. It's
really grunt work. Two, rule clarity. Like,
[11:43] is it Is there a predefined set of rules that
you can apply for that handoff so for for the
[11:47] process? Like, if you understand those rules
and you can apply them, agents can do that, too.
[11:53] Three is the frequency, the volume of
processes, right? If you have process
[11:57] where you have a number of these calls going
or very heavy volume process prime candidates.
[12:02] Um and four, the cost of an error here. How big
is that cost of error, right? Is it something you
[12:07] can live with? If not, you know, you should have
a human in the loop is what we advocate always
[12:12] at the end of the process. So, once you prepare
the balance sheets, financial statements,
[12:16] once you do the recon, you want a reviewer
still have be here reviewer and making sure
[12:22] things are in place. If not, you chat with
the agent and and get the right output again.
[12:27] Um so, in the two by two matrix, you
know, you would see two of these elements,
[12:31] rule clarity and handoff density, on
either axis. Our recommendation is
[12:36] you always transform the ones where there's a
lot of these manual coordination and there's
[12:40] and and there is very clear rules, right?
Because you can make them more deterministic.
[12:44] Prime candidates are accounts uh AP matching,
reconciliations, um close financial close.
[12:52] You know, there's a lot of coordination and
it's high volume is rule clarity, also. Um where
[12:57] you have lower volume, uh sorry, lower handoff
density, uh but you still need some automation,
[13:02] you can just use RPA or you can just use any
automation provider, right? There's not There's
[13:05] not really that much non-determinism. It's very
deterministic in the in the lower bottom quadrant.
[13:11] Because it there is rule clarity. Uh low
volume, you know, and uh low rule clarity,
[13:17] don't even automate. Don't even bother. There
won't be ROI. And then um the last one is uh
[13:23] low rule clarity and higher handoff density.
Those are the ones that you generally want to
[13:29] use AI to augment. You don't want to replace
because this require a lot of human judgment
[13:34] and AI can make mistakes here and the volume,
you know, is generally not that high, right? So,
[13:39] FP&A analysis, uh some exception handling, you
know, you still want your your team to do that.
[13:48] So, let's start wrapping up, right? So, what does
automation to autonomous execution look like?
[13:56] Automation you you would all
have automation tools today.
[14:00] Uh there's RPA bots that you can run, you know,
Python scripts you can run. But essentially,
[14:04] those do what they what they do do really well
is automating a particular task. The moment,
[14:10] you know, the UI changes, they break. The moment
you want to go beyond that task to a series of
[14:14] tasks, you want to apply non-determinism, they
break, right? Autonomous agents is agents running
[14:19] the full workflow end to end. The agents do all
the coordination. The agents do all the execution.
[14:25] The agents loop in humans where required. Agents
start the process. Agents end the process.
[14:30] And humans are in the process reviewing,
right? So, that's the big difference.
[14:38] Uh a plug-in to Adopt here, you know, what we do
at Adopt is we we have the ability to connect with
[14:43] any system. We have 300 plus integrations, many
of them in the finance space. But we have patented
[14:48] technology where we can discover browser APIs and
convert them to API endpoints. So, we connect to
[14:53] all these systems. We have predefined skills and
agents and we bring all of these autonomous agents
[14:58] onto a harness that can execute end to end um and
generate like the outcomes that you need. It's a
[15:05] multi-agent, multi-human collaborative environment
because finance needs that. You have preparers,
[15:10] reviewers, admins, everybody working together
uh in an environment where which is conducive.
[15:17] I'll end with this. So, there's two ways you
can go deeper. Uh one is, you know, you can
[15:21] visit us at our booth here. We're at booth 729 and
we'll give you a customized demo and some swag.
[15:27] Feel free to come and collect
some. Um, or just scan this and uh,
[15:31] scan this QR code and we'll see you um, on
the online with the demo. Thank you all.
