
The shepherd in Luke 15 knew one sheep was missing because he had counted to a hundred, and Jesus tells the story as though counting were the most ordinary part of shepherding. Church software now offers to run that count for the pastor. One platform advertises “at-risk member detection” and a 0 to 10 burnout score for volunteers; another says its AI reads attendance, giving and engagement signals “to flag members who may be drifting.” A drift score is a faithful tool when it sends a shepherd who knows the person’s name to that person’s door, and it turns into surveillance when people are scored, sorted or recognized by camera without being told.
Key Takeaways
- Scripture treats counting the flock as care: the shepherd of Luke 15 noticed the missing sheep because he kept a number.
- Church platforms now sell at-risk flags and volunteer burnout scores, and Churchix sells facial recognition that identifies enrolled members from live cameras.
- Scripture also records a census that went wrong. Joab asked why King David should “delight in this thing,” which places the danger in what the count was for.
- The Shepherd’s Ledger Test asks four questions of any drift score: was the purpose named to members, who sees the score, what the flag triggers, and could the rule be read aloud on Sunday.
- Uganda’s Data Protection and Privacy Act 2019 lists religious beliefs as special personal data and requires a specific purpose and prior notice before collection.
Counting the flock is a pastoral act
Jesus asks his hearers to picture a man with a hundred sheep who loses one, leaves the ninety-nine and goes after the one until he finds it (Luke 15:4). The story runs on arithmetic. A shepherd who had stopped counting would have heard about the lost sheep weeks later, from a neighbor. Proverbs gives the same duty in plain management language: “Be thou diligent to know the state of thy flocks, and look well to thy herds” (Proverbs 27:23).
So a pastor who keeps an attendance register is doing something Scripture commends, and so is the cell leader who sends a WhatsApp message to a member missing two Sundays in a row. In a church of 1,500 people with, say, five staff, memory fails long before love does, and a register is how that staff keeps faith with people it sees once a month.
Software sits on the same line, and the companies building it are doing honest work. A vendor that turns a stack of connection cards into a follow-up list within hours has served a congregation, and that service deserves to be called good first.
What church software now sells
The products have moved from recording attendance to predicting it. Church.IQ bills itself as an agentic AI platform for church management, with “at-risk member detection,” growth forecasting and an AI burnout score on a 0 to 10 scale for people on serving rosters. AlignedAI’s guide to AI church management software says its system analyzes attendance patterns, giving trends and engagement signals so the pastoral team “gets a heads-up early enough to reach out.”
The same company makes its management software free to churches that process gifts through its giving service, at 2.9 percent plus 30 cents per card gift. That is a legitimate business model. It also means the software that flags a drifting member runs on the same rails that carry the member’s tithe.
Facial recognition is the sharper edge. Churchix, built by the company Face-Six, identifies enrolled members from live cameras or recorded video and claims more than 99.83 percent accuracy in controlled environments. The Washington Post covered it in 2015. The Glass Room’s entry on Churchix notes that Face-Six uses similar software in products for casinos, airports and border control.
In 2021 the Wall Street Journal profiled Gloo, a Colorado company working with more than 30,000 churches, about 10 percent of US congregations, and holding data on some 245 million Americans, according to MinistryWatch’s summary of that profile. The Christian Post’s account of the same reporting describes a Gloo report for a church near Kansas City that predicted 25 percent of marriages within five miles of the church might be near divorce and 26 percent of people were at risk of opioid addiction. Gloo told the Journal it had stopped using mental health data after the paper began reporting on its work, and it declined to say where its data came from.
Every usage figure in this section is American. I looked for a published count of churches in Uganda, Tanzania or Ethiopia using predictive attendance tools or facial recognition and found none, so read this as a description of products already on sale to any church with a card and a login.
The strongest case for the drift score
The vendors’ best argument is pastoral, and it should be heard at full strength. AlignedAI puts it plainly: by the time a pastor notices a three-month absence, “they’ve already emotionally checked out.” Randy Frazee, lead pastor of the church that used Gloo, gave the theological version, as the Christian Post reported it: “The church is committed to going out at whatever cost to find that one lost sheep that needs help.”
Concede what is true here. Large churches lose people quietly, and by the vendor’s account the loss is usually noticed late. A member who stops coming after a bereavement or a job loss can drift for weeks before anyone asks after her. An early flag that puts her name in front of a cell leader in week three, instead of week twelve, is mercy delivered on time. Given a choice between a model that catches her and a register that sits unread in the office, the model wins, and Frazee is right about the duty to go out and look.
The argument covers the first half of the parable. The second half is about who goes, and what the sheep know about the one who comes for them.
David’s census and Ezekiel’s shepherds
Scripture records counts that were commanded and a count that was judged. God tells Moses to number Israel in Numbers 1. In 2 Samuel 24, David orders a census, and his own general objects before the counting begins: “why doth my lord the king delight in this thing?” (2 Samuel 24:3). The text leaves the full reason for the judgment unstated, but Joab’s question places the danger in the king’s motive. The same arithmetic served Moses as an act of order and served David as a mirror for his own strength.
Ezekiel 34 names what a count becomes when the shepherd is its beneficiary. Israel’s shepherds are told, “Ye eat the fat, and ye clothe you with the wool” (Ezekiel 34:3), and they ruled the flock “with force and with cruelty” (34:4). A drift score that blends attendance with giving trends has to answer Ezekiel directly. When a member’s giving drops, does the flag send someone to ask about her job, or does it tell the finance team which households to chase before the building fund closes?
Jesus answers Ezekiel in John 10. The good shepherd “calleth his own sheep by name, and leadeth them out” (John 10:3), and he says he knows his sheep and is “known of mine” (10:14). The knowing runs both ways. A facial recognition camera reverses the direction: the church knows the face, and the person in the pew is the last to find out.
Psalm 139 describes the one observer whose complete knowledge is safe, because it belongs to the one who made us. Even there, the psalmist invites it: “Search me, O God, and know my heart” (Psalm 139:23). The searching is something the searched person asks for. A church that wants to know its people well has the same door open to it. It can ask them.
The Shepherd’s Ledger Test
Before any drift score, at-risk report or camera goes live, put four questions to it. I call this the Shepherd’s Ledger Test, because a ledger is a record someone has to answer for.
- Was the purpose named to the people being counted? Data kept for pastoral follow-up and run through a model that also weighs giving now has a second purpose, and members deserve to hear it from the pulpit before the model runs.
- Who sees the score? Keep the circle to the member’s own cell leader and pastor. A drift score visible to the finance office, a leadership selection committee or a partner ministry has left the shepherd’s hands.
- What does the flag trigger? A flag should end at a doorstep, never in a spreadsheet. Every alert needs a named person who will call or visit within a set number of days, and a record of whether the call happened. If flags pile up unanswered, switch the feature off until there are people to answer them.
- Could you read the rule aloud on Sunday? Gloo stopped using mental health data once a newspaper started asking about it. That is the test run backward. If a practice would have to change the day the congregation learned of it, change it now.
Facial recognition struggles with questions one and four by design, since the camera works best when people forget it is there. Predictive flags can pass all four if the church tells people and staffs the follow-up. This sits close to the line drawn in AI pastoral care and the limits of triage: software may sort the list, and a person carries the care. The wider duty over the records themselves is set out in Your Church Database Is a Sacred Trust.
As an illustration, picture an executive pastor at a 1,500-member Kampala church whose vendor has just switched on an at-risk report. The test fits on one page. Purpose: pastoral follow-up only, announced on two Sundays. Viewers: cell leaders, for their own cells. Trigger: a phone call within five days, logged. Giving data: excluded from the model. If any line is hard to write, the feature stays off.
What Uganda’s data law already asks of a church
Section 9 of Uganda’s Data Protection and Privacy Act 2019 lists data relating to religious or philosophical beliefs among the categories of special personal data, alongside health and financial information, and restricts processing it. It then permits processing for the legitimate activities of a body that exists for religious purposes, where the data relates to its members or people in regular contact with it, and where any disclosure to a third party has the person’s consent. An attendance list records religious practice about as directly as data can, so on my reading the register falls inside section 9. Have a lawyer check that reading against your systems and vendor contracts.
Two more sections bear directly on prediction. Section 12 requires data to be collected for a lawful purpose that is “specific, explicitly defined” and related to the collector’s functions. Section 13 requires telling people, before collection, what kind of data is gathered, who is gathering it and why. The Act also gives data subjects rights in relation to automated decision-taking, in section 27. The Personal Data Protection Office, under NITA-U, became operational in August 2021.
The first enforcement case is a lesson about purpose. On 10 July 2025 Uganda recorded its first conviction under the Act: the director of a microfinance firm running a loan app was fined UGX 300,000 for failing to register with the PDPO. A second count, over borrower data collected for loan processing and reused in a threatening video, was settled with compensation to the complainant, according to a Captain Compliance summary of the case. Swap loan processing for pastoral follow-up, and the video for a targeted giving appeal, and you have the risk a church runs when one dataset quietly takes on a second job. Entities that process personal data must register with the PDPO, and a church office with a member database fits that description.
This section reads the Ugandan Act only. A church in Dar es Salaam or Addis Ababa should have its own national statute read with the same four questions in hand.
Frequently Asked Questions
Is it wrong for a church to track attendance?
Scripture commends it. The shepherd in Luke 15 found the lost sheep because he counted, and Proverbs 27:23 tells owners to know the state of their flocks. The danger lies in what the count is for and who it serves. A register that sends a cell leader to a missing member’s door is pastoral care.
Should a church use facial recognition to take attendance?
It is hard to justify. Facial recognition works best when people forget the camera is there, which conflicts with telling members what is collected and why. In Uganda that notice duty sits in section 13 of the Data Protection and Privacy Act. QR check-in or a signed register meets the same need openly.
Can a church combine attendance data with giving data in one score?
It can, and that is where most of the risk sits. A drop in giving may signal a job loss a pastor should ask about. The same signal can steer fundraising toward particular households. If you combine them, name that purpose to members first and keep the score away from the finance team.
Does Uganda’s data protection law apply to churches?
On a plain reading, yes. Section 9 of the Data Protection and Privacy Act 2019 treats religious belief as special personal data and permits a religious body to process it for members, with consent needed for disclosure to third parties. Sections 12 and 13 require a specific purpose and prior notice. Have a lawyer check the details.
What is the Shepherd’s Ledger Test?
It is a four-question check for any church drift score or attendance camera. Was the purpose named to the people counted? Who sees the score? Does the flag trigger a named person calling or visiting within set days? Could the rule be read aloud on Sunday without changing it? A tool that passes all four is serving pastoral care.
Next Step
This week, before your vendor’s next release switches on another feature, write the four Ledger answers on one page and read them to your elders. If you are setting data and AI policy for a denomination or a network of congregations, book a strategy call with Joshua. An hour on the policy now costs far less than a congregation’s trust spent later.
Related Reading
- Your Church Database Is a Sacred Trust
- AI Pastoral Care: Triage Yes, Shepherding Never
- Algorithmic Bias Is a Neighbor-Love Issue for Africa
- What Christian Companies Should Ban, Require, and Audit in AI Tools
Sources and Evidence
- More than 30K churches using big data from tech firm Gloo to target new members (The Christian Post, 27 Dec 2021): Gloo’s report predicting 25% of nearby marriages near divorce and 26% at opioid risk; Randy Frazee’s lost-sheep quote; Gloo dropping mental health data after WSJ reporting and declining to name its data sources.
- Gloo Mines Big Data to Connect Seekers with Churches (MinistryWatch): WSJ 2021 profile figures of 30,000+ churches, about 10% of US congregations, data on some 245 million Americans; Gloo based in Boulder, Colorado.
- Churchix: Facial Recognition Software for Church Attendance: identification of enrolled members from live cameras or recorded video; the claimed 99.83% accuracy in controlled environments.
- Skipping church? Facial recognition software could be tracking you (The Washington Post, 24 Jul 2015): 2015 coverage of Churchix and how it matches faces against a photo database.
- Facial Recognition for Churches (The Glass Room): Face-Six, the company behind Churchix, uses similar software in products for casinos, airports and border control.
- Church.IQ: The AI-First Church Management Platform: at-risk member detection, growth trend forecasting, an AI burnout score on a 0 to 10 scale for roster members.
- AI Church Management Software: What’s New in 2026 (AlignedAI): predicting who is slipping away from attendance, giving and engagement signals; the three-month quote; free software tied to its giving service at 2.9% plus 30 cents per card gift.
- Uganda Data Protection and Privacy Act, 2019 (ULII / Laws.Africa): sections 9, 12 and 13 text; section 27 on automated decision-taking (table of contents).
- Data protection laws in Uganda (DLA Piper): section 9 special personal data definition; PDPO under NITA-U operational August 2021.
- Uganda’s Data Protection Law: First Ever Fine, Mandates, and Comparison to POPIA and GDPR (Captain Compliance, 7 Aug 2025): first conviction on 10 July 2025, UGX 300,000 fine, purpose-limitation count over a loan app’s shaming video, registration requirement.
