Blog

<figure id="figure-i079oxjmdnmvzobbbg"><img id="i079oxjmdnmvzobbbg" src="https://cdn.durable.co/blocks/22s9Kc60Bu4Q4irlvbwCb8oMGEYyHnvKF9iyNmyr2YH8qgDkJD49m3iKkRFUOlJE.png" alt="Navy graphic reading meetings, not emails sent, with bars shrinking from sends to meetings" title="Navy graphic reading meetings, not emails sent, with bars shrinking from sends to meetings" class="object-cover bg-gray-100 " style="aspect-ratio: 3/2; object-position: center center;" data-corners="default" data-aspect="3:2" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure><p>IT asset disposition and e-waste companies live on a steady flow of pickups. That makes outreach important, and it makes volume tempting. Send more emails, reach more companies, book more loads.</p><p>It rarely works that way. In our work with ITAD and e-waste operators, the numbers that predicted revenue were never the send counts. They were replies, meetings, and loads actually picked up.</p><h2>Why volume misleads</h2><p>A large send number feels like progress. But every unverified address that bounces, and every message that lands in spam, makes the next message less likely to arrive. Push volume too hard on a new or shared sending domain and you can damage the channel you depend on.</p><p>Meanwhile the report says thousands of emails went out, and nobody can point to the meetings they produced.</p><h2>What we measure instead</h2><p>1. Replies: a real person answered, positive or not.</p><p>2. Meetings and calls booked: the conversation moved forward.</p><p>3. Pickups and closed deals: the only number that pays.</p><p>If a change raises volume without raising any of these, it is the wrong change.</p><h2>What we learned</h2><p>Verify before you claim it is verified. A lead list described as checked is not checked until the results file exists. We have seen lists labeled verified that never were.</p><p>Protect the sending domain first. Warm a dedicated domain before sending at scale. Until then, a short weekly list of verified leads for the sales team to call can outperform a large email campaign.</p><p>Keep each client's data separate. Lists, templates, and logs for one operator never mix with another's.</p><h2>Where automation actually helps</h2><p>Automation is useful for the slow parts: finding companies that match your service area, checking contact details, drafting follow ups for a person to approve, and logging every outcome so the numbers above are real. It is not useful as a way to send more of something that is not working.</p><h2>A one week audit</h2><p>For one week, write down three numbers from your current outreach: messages sent, replies received, and meetings booked. If the gap between the first and the other two is large, the fix is not more sending.</p><p>If you run an ITAD or e-waste operation and want help with that audit, email <a href="mailto:von@descieuxdigital.com">von@descieuxdigital.com</a></p>

ITAD Outreach: Count Meetings, Not Emails Sent

<figure id="figure-9905brlltmvzo6b5l"><img id="9905brlltmvzo6b5l" src="https://cdn.durable.co/blocks/32zHe7Trovk5oI1ropuxullIeBbGiUMHWJtRLc6Q4nxbRaxOTgxfWSv0KWoVwaZY.png" alt="Navy graphic reading 7:00 AM, the close package is ready, with a clock" title="Navy graphic reading 7:00 AM, the close package is ready, with a clock" class="object-cover bg-gray-100 " style="aspect-ratio: 3/2; object-position: center center;" data-corners="default" data-aspect="3:2" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure><p>In a lot of finance teams, the first hour of the day goes to assembly. Someone pulls the bank extract. Someone updates the dashboard. Someone drafts the journal entries and builds the list of items that need attention. Only after all of that does anyone start the work that actually needs a person.</p><p>That assembly is a good job for a machine.</p><h2>What arrives overnight</h2><p>For one client, we built a nightly job that does the assembly while the office is closed. By the time the team logs in, four things are waiting:</p><p>1. An executive dashboard with the prior day's numbers, ready to share.</p><p>2. A draft journal entry for the routine items, prepared but not posted.</p><p>3. An exception list of the items that did not tie, each with the reason.</p><p>4. A snapshot of the data the package was built from, so any number can be checked later.</p><p>The team starts the day reviewing, not rebuilding.</p><h2>People still approve</h2><p>Nothing in the package posts itself. The journal entry is a draft. The exception list is a to do list, not a set of decisions. The controller and the team sign off, the same way they would on work prepared by a junior analyst.</p><p>The difference is that the preparation is consistent, it happens every night, and it is ready before anyone arrives.</p><h2>When something is missing</h2><p>If the night's data is incomplete, the job does not quietly produce a partial package. It stops and reports what was missing. A morning with a clear warning is better than a morning with a dashboard nobody should trust.</p><h2>What it takes to start</h2><p>You do not need to automate the whole close. Start with one piece:</p><p>1. Pick the report your team rebuilds most often. Daily cash, a weekly flash, a recurring reconciliation.</p><p>2. Make sure the source extracts are consistent. Same format, same schedule.</p><p>3. Name the reviewer. Someone owns the output every morning.</p><p>Once one package runs reliably, adding the next one is much faster than the first. Because we build on the exports you already have, the first version is usually measured in weeks.</p><p>If your team spends its mornings assembling reports, email <a href="mailto:von@descieuxdigital.com">von@descieuxdigital.com</a></p>

The 7 AM Close Package: What Lands Before Coffee

<figure id="figure-fem400wzt8mvznwfte"><img id="fem400wzt8mvznwfte" src="https://cdn.durable.co/blocks/169e9wkIYIodNqgHzVRjsedv3tVIq8WDXxMj7nUdbOdfFROCnTmQC0tz9I7i5tKt.png" alt="Navy graphic reading complete, partial, refused, with green, amber and red status tags" title="Navy graphic reading complete, partial, refused, with green, amber and red status tags" class="object-cover bg-gray-100 " style="aspect-ratio: 3/2; object-position: undefined undefined;" data-corners="default" data-aspect="3:2" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure><p>The most dangerous financial report is not the one with an obvious error. It is the one that looks clean and is quietly wrong.</p><p>That usually happens the same way. A file arrives with a month missing, or a segment that did not export, or a column that came through blank. The reporting system does not stop. It treats the missing values as zero, totals everything up, and produces a report that looks finished.</p><p>Nobody notices until someone asks why a number moved.</p><h2>Zero is not the same as missing</h2><p>A zero means "we checked and there was nothing." A blank means "we do not know." Treating the second as the first is how a report ends up understating expenses, overstating margin, or showing a trend that is really just a gap in the data.</p><p>In the finance systems we build, missing data is never filled in silently. Every report carries a status:</p><p>1. Complete: every input the report needs was present and checked.</p><p>2. Partial: some inputs were missing, and the report names exactly which ones.</p><p>3. Refused: too much was missing to produce numbers anyone should rely on, so the system does not produce them.</p><h2>What this looks like in practice</h2><p>For one client, a nightly job assembles the finance team's morning package. If one segment of the data is absent, the job does not write a package with that segment set to zero. It stops, records what was missing, and tells the team before anyone opens a stale or incomplete dashboard.</p><p>That can feel inconvenient the first time it happens. It is far cheaper than a board slide built on a number that was never real.</p><h2>Why auditors like it</h2><p>When every report says what it was built from, and refuses to guess, the trail is easy to follow. Any number can be traced back to the files and rows that produced it. Gaps are documented at the moment they happen, not discovered months later.</p><h2>How to test your current reports</h2><p>Try this with any recurring report your team relies on:</p><p>1. Take last month's input file and delete one week of rows.</p><p>2. Run the report again.</p><p>3. See whether it warns you, or simply shows smaller numbers.</p><p>If it shows smaller numbers without a word, it will do the same thing the next time an export fails.</p><p>If you want help building reports that say what they do not know, email <a href="mailto:von@descieuxdigital.com" data-mce-href="mailto:von@descieuxdigital.com">von@descieuxdigital.com</a></p>

When the Data Is Incomplete, the Report Should Say So

<figure id="figure-6nknps6bknmvzngaul"><img id="6nknps6bknmvzngaul" src="https://cdn.durable.co/blocks/36L5amxYEIL9qZvJaLsvKcx6bmpV5MPbCb95xtovS4fOCHbehd99mLWwtuIPmCAt.png" alt="Navy graphic reading code computes, the AI explains, with a calculated total" title="Navy graphic reading code computes, the AI explains, with a calculated total" class="object-cover bg-gray-100 " style="aspect-ratio: 3/2; object-position: undefined undefined;" data-corners="default" data-aspect="3:2" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure><p>Ask a language model to add up a column of numbers and it will usually give you an answer. It will sound confident. Sometimes it will be wrong, and nothing in the answer will tell you which time that was.</p><p>For most uses that is a minor problem. For a finance team, it is the whole problem. A number in a board report has to be right, and it has to be right the same way every time someone runs it.</p><p>So we follow one rule in every finance system we build: code computes, the AI explains.</p><h2>What that rule means in practice</h2><p>Every figure that matters (revenue, expenses, balances, variances, forecasts) is calculated by deterministic code. The same file goes in, the same numbers come out, every run. Headline totals are summed from the detail underneath them, never copied from text a model wrote.</p><p>The AI model gets the finished numbers and does what it is good at: it writes the summary, points out what changed, and puts the results in plain language for the person reading the report.</p><p>If the model's explanation and the computed numbers ever disagree, the computed numbers win, and the report says so.</p><h2>A real example of why this matters</h2><p>While building our own finance reporting tool, we found a bug in how one kind of row was being read. Cost rows that were entered as positive numbers were being counted as revenue. On the affected data, reported revenue was overstated by about 20 percent.</p><p>It had gone unnoticed because the sample data used in testing happened to follow a convention that hid it. The fix was simple once it was found. What found it was not a model reading the output and deciding it looked off. It was deterministic checks comparing totals that should have agreed and did not.</p><p>That bug is now covered by tests that run on every change. That is the kind of safeguard you can only build when the math lives in code you can inspect.</p><h2>Why this matters when you choose a vendor</h2><p>Plenty of tools now promise AI for finance. Some of them let the model produce the numbers directly. That can look impressive in a demo and still fail quietly on your data.</p><p>Before you trust any AI finance product, ask three questions:</p><p>1. Where are the numbers calculated? In code you can audit, or by the model?</p><p>2. If I run the same file twice, do I get the same totals? If the answer is "usually," keep looking.</p><p>3. What happens when data is missing? A good system tells you what it could not compute. A bad one fills the gap and moves on.</p><h2>The short version</h2><p>AI is very good at explaining financial results. It should not be the thing producing them. Keep the math in code, keep the explanation in the model, and you get reports that are both readable and correct.</p><p>If you are evaluating AI for your finance team and want a second opinion on how a tool handles its numbers, email von@descieuxdigital.com.</p>

Why Our AI Never Does the Math

<figure id="figure-jevm7a6ghjmvzn4t4p"><img id="jevm7a6ghjmvzn4t4p" src="https://cdn.durable.co/blocks/388t2z2AMwJUTudTgZeOskQW9I1al73Sn1r6Pp6uVW7udnYoxHAbWRYMoWF7Wd3U.png" alt="Navy graphic reading 144 days, tied to the penny, with ledger lines checked off" title="Navy graphic reading 144 days, tied to the penny, with ledger lines checked off" class="object-cover bg-gray-100 " style="aspect-ratio: 3/2; object-position: undefined undefined;" data-corners="default" data-aspect="3:2" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure><p>Most finance teams know the feeling. The bank shows one number, the general ledger shows another, and someone spends the next two days finding out why. By the time the difference is explained, the month has moved on and the next batch of sweeps is already waiting.</p><p>Reconciliation is not hard work. It is slow work. It is the same comparison done hundreds of times, and the one line that matters is buried among the ones that match. That makes it a good place to start with automation.</p><h2>What we built</h2><p>For a regional financial services firm, we built a reconciliation engine that compares every daily sweep against the ledger. It reads the bank extract and the GL detail, matches the two line by line, and produces one output: a short list of the items that do not tie.</p><p>On that firm's data, the engine reconciled 144 days in a row to the penny. Not close. Not within tolerance. Exact.</p><p>In another engagement, the same approach matched more than 90 reconciliation lines automatically and isolated a six-figure exception. A manual process would likely have taken days to surface it.</p><h2>What changes for the controller</h2><p>The controller's job does not go away. It gets narrower and more useful.</p><p>Before, the team reviewed everything to find the few things that were wrong. Now they review only the exception list. Each exception arrives with the bank line, the ledger line, and the reason they did not match, so the question shifts from "where is the problem?" to "what do we do about it?"</p><p>That is the part that needs judgment, and it is the part people should spend their time on.</p><h2>Why exact matters</h2><p>A reconciliation that is "almost right" is still a reconciliation someone has to check by hand. The point of automating it is that the matched days can be trusted without a second look. That only works if the matching is deterministic: the same inputs give the same result every time, and every match can be traced to the two lines that produced it.</p><p>We do not let an AI model decide whether two numbers agree. Code does the matching. Where AI helps is in explaining the exceptions in plain language and drafting the notes your team would otherwise write.</p><h2>What to have ready before you automate</h2><p>You do not need a new system to start. You need three things:</p><p>1. A clean bank extract in a consistent format, pulled on a regular schedule.</p><p>2. GL detail for the same accounts and dates, not a summary.</p><p>3. A named owner who reviews the exception list and decides what happens next.</p><p>With those in place, the first version of a daily reconciliation can usually be running within weeks, not quarters. That speed is the main difference from a large consulting firm engagement: we build on the systems and exports you already have.</p><h2>The short version</h2><p>Let the machine do the matching. Let your team handle the exceptions. And insist that every matched day ties exactly, so nobody has to check it twice.</p><p>If your team is still reconciling by hand, email <a href="mailto:von@descieuxdigital.com" data-mce-href="mailto:von@descieuxdigital.com">von@descieuxdigital.com</a> and we will look at one month of your data with you.</p>

Bank Reconciliation That Ties to the Penny, Every Day

Diverse IT team collaborating on system migration and cloud infrastructure implementation with technical diagrams and monitoring displays

<h2>Understanding Cloud Migration</h2>
<p>Cloud migration is the process of moving data, applications, or other business elements to a cloud computing environment. For businesses in Los Angeles, a successful cloud migration can lead to enhanced scalability, improved performance, and cost savings. However, migrating to the cloud requires careful planning and execution.</p>
<figure id="figure-0v0akbvmpmrp8umpe"><img id="0v0akbvmpmrp8umpe" src="https://rjdavx8ozyznxeyh.public.blob.vercel-storage.com/production/websites/infinite-images/e5cae44f-f980-4a97-9592-4be72032a99c-YlvQBbw9Y21XdHAZmZbjDAZNGKQqw8.png" alt="cloud migration" class="object-cover bg-gray-100" style="aspect-ratio: 16/9"></figure>

<h3>Assessing Your Current Infrastructure</h3>
<p>Before migrating to the cloud, it's crucial to assess your current IT infrastructure. This involves understanding what applications and data you are running, how they interact, and identifying any dependencies. Conducting a thorough assessment will help you determine the best cloud solutions for your needs.</p>
<p>Consider using tools that can automatically map your existing infrastructure. Such tools can provide insights into potential challenges and help you develop a tailored migration strategy.</p>

<h2>Choosing the Right Cloud Provider</h2>
<p>One of the most important decisions you'll make during your cloud migration is selecting the right cloud provider. Consider factors such as cost, security, scalability, and support. Major providers like AWS, Google Cloud, and Microsoft Azure offer a variety of services, but the best choice depends on your specific requirements.</p>
<figure id="figure-etzx0w271emrp8umpe"><img id="etzx0w271emrp8umpe" src="https://rjdavx8ozyznxeyh.public.blob.vercel-storage.com/production/websites/infinite-images/4f1176d9-6f01-49ca-af58-3a421fbfde1b-f2ScHbsQhtYwZemOxaU06nHZdnHD9J.png" alt="cloud provider comparison" class="object-cover bg-gray-100" style="aspect-ratio: 16/9"></figure>

<h3>Planning Your Migration Strategy</h3>
<p>Once you've chosen a provider, it's time to develop a detailed migration strategy. This involves deciding which applications to move first, setting timelines, and preparing your team for the transition. A staged approach often works best, allowing you to address any issues before fully committing to the cloud.</p>
<p>Engage with experts who can guide you through the process and ensure best practices are followed.</p>

<h2>Ensuring Data Security and Compliance</h2>
<p>Data security is a top priority during cloud migration. You must ensure that all sensitive information is protected and complies with relevant regulations. In Los Angeles, businesses must pay attention to laws such as the California Consumer Privacy Act (CCPA).</p>
<figure id="figure-iatkpt1zw6mrp8umpe"><img id="iatkpt1zw6mrp8umpe" src="https://rjdavx8ozyznxeyh.public.blob.vercel-storage.com/production/websites/infinite-images/dba70514-b76f-4bef-b03b-abe9b5936bdc-NIuY8tVCTLliulqjwOyDoQBQnO58og.png" alt="data security" class="object-cover bg-gray-100" style="aspect-ratio: 16/9"></figure>

<h3>Optimizing Costs</h3>
<p>Cloud migration can be a cost-effective solution if managed properly. Start by analyzing your current expenses and projecting potential savings in the cloud. Many providers offer cost management tools to help you track and optimize spending.</p>
<p>Consider implementing policies that prevent unnecessary resource usage, such as shutting down non-essential instances during off-peak hours.</p>

<h2>Training and Supporting Your Team</h2>
<p>Your team’s understanding of cloud technologies is vital for a successful migration. Provide training sessions and workshops to enhance their skills. This not only eases the transition but also empowers your team to leverage the full capabilities of cloud services.</p>
<figure id="figure-mwrw12bm6mrp8umpe"><img id="mwrw12bm6mrp8umpe" src="https://rjdavx8ozyznxeyh.public.blob.vercel-storage.com/production/websites/infinite-images/36938617-f7fe-4410-9a9a-dd792351637e-JX8Riko6zNm1IpVrxxEfPUvtNarmM1.png" alt="team training" class="object-cover bg-gray-100" style="aspect-ratio: 16/9"></figure>

<h3>Monitoring and Optimization Post-Migration</h3>
<p>After migrating, continuous monitoring is key to ensuring your systems are running smoothly. Utilize monitoring tools to track performance and address any issues promptly. Regularly review your cloud setup to optimize and adapt to changing business needs.</p>
<p>By following these expert tips, businesses in Los Angeles can achieve a seamless and successful cloud migration, unlocking new opportunities for growth and innovation.</p>